Rebuilding CDR3 is still very hard

[douweschulte]

Here is a very nice example where a couple of problems are visible:

• Too short overhang for the last two reads
• Slightly misaligned reads, asking the question if a multiple sequence alignment is warranted
• With this level of rebuilding would an additional step between TM and REC be needed, essentially being REC but then use the germline with the reconstituted CDR3 as template (quite similar to Template Refine #208)

[MengTingHe2023]

I've also noticed this issue; the CDR3 of the recombined sequence generated with Stitch always seems to be much longer than the reference. And raising the CutoffScore doesn't seem to resolve the issue either.

[douweschulte]

Can you show me a picture of how your recombination looks (with the peptides visible for both sides IGHV + IGHJ) that way I can help you think if there is a way to get Stitch to recombine it better?