You do want to get the two sections to join as flat as possible. If they don't match up flat, when you sand, you may go through the upper layer of ply and it will not look as good. If you already have a mismatched joint, you can heat the joint to soften the epoxy and reseat it. If your work surface is not real sturdy, I'd use the floor. I also test join the joints dry just to be sure there are no interferences. Just be careful separating them back out. Don't thicken the epoxy too much for this joint as there isn't much space and you want to be sure the edges are wetted out. And yes, you will need lots of cups, and gloves.

Check out Nick Schade's video regarding puzzle joints.  He does not use epoxy to hold the joints together.  Instead, he uses dots of super glue which allows you to very precisely get the panels flush.  The glue is strong enough to hold the joint together during boat construction.  When the boat gets glass, enough epoxy gets into the joint to make it full strength.  This is standard procedure in the Petrel and Petrel Play manuals.  Since building my Petrel Play, I have used this technique on every puzzle joint with good success.

   I'll see how they look later today.  I realized later on that the reason they wouldn't sit flush was from slight curves in the wood, and that they would (hopefully) lay flat with weight on them.  Each joint is between two totally flat surfaces and has a cinder block on top.   I mixed the second small batch of thickened epoxy a little thinner, and that seemed to help as well.  Live and learn I gues.

I'm planning to paint the bottom and sides with only the top plank finished bright.  I haven't glued that one yet.  The interior will be painted as well, with only the inwhales, hatches, and smaller bits finished bright.  So hopefully I have a little room for learning.  Still buying a heat gun today though...

   I checked the joint I had the most trouble with.  Some tabs of the puzzle joint are off by probably .5mm and would probably require too much sanding.  It's the bottom panel which is getting painted on both sides, but it would still need to be faired or re-seated.  Hopefully the rest are within spec.  I'm glad I didn't try to glue them all in one go like I'd planned.

   I am no expert on puzzle joints but you may be able to use a heat gun to loosen the misaligned joint sufficiently in order to align it.   

   On close inspection, 3 of the 4 joints I glued up are workable, one of them is pretty bad.  No way it could be finished without paint (luckily, it will be painted either way).  On the worst area of the bad joint I might have to sand halfway through a ply.  I feel like that is too much.   I started to work on it with the heat gun, but made no progress and got the sinking feeling I would make things worse if I kept going, so I stopped.  What are my options for smoothing the joint if I don't want to redo it and I intend to paint it anyway? 

I thought my idea to press the joints between slate floor tiles was a good one, but I need to make the fit flat in the first place.  Also, I made the mix too thick and probably used too much.  Plenty of joints left to go, I'll get it eventually.

halfway through the upper layer, not the whole ply...  I wish we could edit our posts.   

   Really, the heat gun should work, go slow to allow heat to sink in, keep some shear pressure on the joint (maybe just prop one side of the joint up on a scrap board and push down on the other side with your fingers, but wearing some leather gloves to avoid burnt fingers) so that it will move as the epoxy softens, and do avoid burning the wood. Maybe flip the plank half way throught the heating to get both sides nice and warm before trying to press it apart. I'd give it another shot.

If you do elect to keep the joint as is, you might need to fair it back as much as 6 inches or more from the joint with epoxy filler on both sides of the plank, each on the low side of the joint (which means the filler will be offset on opposite sides of th plank) then sand as flat as possible over the whole width of the joint. The flatter you can fair the wet epoxy, the less subsequent time-consuming and difficult sanding you'll need to do.  Micro-balloons used in the filler/fairing epoxy will make sanding much easier. Even if you don't have balloons now, you might order some for future use - you'll find the opportunity to use them at some point. I believe sanding epoxy with wood flour is slightly easier than pure epoxy, so if that's all you have, and you are painting anyway, I say use it. Sanding epoxy with cellofill seems hardest of all to me (except for silica, which is even harder, but shouldn't be a consideration here), so don't use either of these for filler in fairing/filling epoxy mixes.

Try a nice long (at least 12 inch) fairing board as a sanding block, and try manual sanding. Things will get wavy if you just run over the joint area with your palm sander.

The usual description CLC/John Harris gives for thickened epoxy applications (cellofill) is "mustard" consistency.  Normal for filet epoxy applications (wood flour) is "peanut butter" consistency.  I think it is pretty accurate guidance.

   One thing I didn't make clear - unless things are looking good otherwise during "re-seating," that wouldn't be my first choice. I think it would be better to disassemble the joint with heat, scrape while warm, let things harden, clean, sand and flatten the planks and clean the edges of the puzzle joint, then re-assemble from scratch.

   Thank you all very much for the advice.  I'll make a decision about how to handle it.  I am so happy that I've only joined the #1 planks, and the bottom planks.  6 joints in total (I mistakenly mentioned 4 earlier).  Only one of them is atrocious, one is borderline, and the rest should be within reason.  I am not doing anymore joining until I sort this, and will be far more careful going forward.  I know what most of my mistakes were by now.  It's a good thing I have a climate controlled shop and all winter to work without feeling rushed.  I can still enjoy the process even during this self-induced drama.  

   You might want to have a look at Nick Schade's recent (5 days ago) YouTube video on assembling puzzle joints:  CA glue only, no epoxy

   For your future puzzle joints you may wish to adopt the following: Once I've pre-fitted the puzzle joints, without goop, and done any sanding necessary to get them to fit, I'll then swab the edges of each puzzle joint with thickened goop.  Somewhere around mustard consistency is fine.

Once I've fitted the joints together, I'll scrape away any excess.  I'll then put wax paper on top of the joint, and wax paper below the joint.  The epoxy is still wet.  The wax paper should completely encompass the entire joint.

I'll then place a piece of 1/2" thick or more plywood, on top of the joint, on top of the wax paper, such that the ends of that plywood go a few inches beyond the edge side of the puzzle joint, on both ends, and place another piece of plywood on the other side of the joint, below the wax paper.  

I'll then use clamps, or screws, to tighten both pieces of plywood together, so as to basically "sandwich" the puzzle joint tight.  I might also place blocks of wood downwind and upwind of the puzzle joint, underneath, so as to keep everything level.  The sandwich helps to ensure that the result is a level joint.  

The next day I'll remove the sandwich, sand the puzzle joint, and add any thickened goop into places where there are gaps.  Later I'll sand the joint again.  

   Thanks again everyone for the advice.  I am almost out of the woods.  Lots of sanding, and one round of fairing on the worst spots and all is almost well.  I have a couple spots that need another round of fairing, and then I'll be moving on to the joining the next planks.   As mentioned, these first joints will all be under glass and painted, so it shouldn't be a big deal other than time served.  I have learned a bit about epoxy, sanding, and fairing.

To recap for anyone who may want to avoid fiding themselves in a similar situation, what I did wrong was:

1) Not dry fiting the parts to make sure they would lay flush.   I beat the snot out of the stern bottom panel with a rubber mallet but two of the puzzle tabs would not lay flat even when the rest of the joint was flush.  I should have taken the joint apart and removed all of the epoxy I had applied to the mating surfaces and fixed the problem, but I thought it would be within sanding range.  Not so.  When it was fully cured, it was worse than it was before hand.  The unlevel surface created a space between the joint and the plastic for epoxy to pool on the bottom side.

2) I used too much epoxy.  My main concern was to fully coat the mating surfaces, thinking I would scrape the excess off after the joint was set.  Now that I know about the CA glue method, I might have wanted to use it since these first joints will all be glassed on both sides.  

3) I may have mixed the batch too thick.  It was difficult to coat the mating surfaces without making a mess.  I will try a slightly thinner mix, and will use a smaller, more precise brush to put the epoxy where it belongs and not all over the place.

4) I was rushing.  I had it in my head that the epoxy wouldn't last long in the cup before going off, so I needed to work fast.  In the future I will err toward wasting epoxy instead of messy joints.

In the end, it all should turn out ok.  Most of the joints may even look fine once glassed.  The stern bottom joint will have visible fairing compound.  I bought a 16" sanding board which will get a lot of use regardless of whether or not I bugger up anymore joints.  I have an extra quart of micro balloons on hand now, too, as well as a heat gun.  Speaking of that, I never was able to loosen the worst joint with heat.  It was the bottom panel, so maybe the thicker ply and larger surface area wouldn't take the heat rapidly enough?  I didn't want to burn the wood, and I just could not get it to budge.  The best I could do was scrape of the excess epoxy, which was helpful.  

The next joints will go much better.  Even though they will have paint, I'm going to treat them as if they will be visible.