Hi @lca123,
So the way I do rough overlap calculations are, expected amplicon length = 805R-541F = 264bp. Overlap = 2x250 cycles - 264 = 236bp! On second thought, this doesn’t look like a typical V3-V4 region, in fact based on the 541F this would be more V4 only. Can you double check to make sure that is indeed the right forward primer, not something like 341F? Especially since you mentioned, your merged amplicons are ~428 bp. Double check and let us know, if it is indeed 541F then we might be dealing with a whole other issue of Forward reads reaching the other side of the Reverse primers etc.
For now I’m going to assume it is 341F-805F 
So that would make your amplicon length 805-541 = ~464bp. Overlap would be 2x250 - 464 = 36bp. Notice how I don’t include the primer trimming in these calculations since those come off the 5’ of our reads and overlap occurs at the 3’, so won’t be affected. With your current truncating parameters you are truncating (250-224) + (250-223) = 53bp which is indeed more than the 36bp overlap. So clearly there’s probably some merging issues. Might be worth doing the forward reads only if you can’t find enough overlap with this primer set.
Edit: Where are these merging statistics you reported from? They seem different than the DADA2 output from earlier.