Improve Ukkonen boundaries about edlib HOT 2 OPEN

I found optimal boundaries, one for first block is working great, however one for last block does not work and I do not know why. I examined it for very long but could not find the problem. However it does not improve much compared to current non-optimal boundary so it does not affect computation speed too much, however it is worrying that it does not work because I found that boundary in same way like boundary for first block. Maybe everything is wrong? I do not believe so, I think maybe boundary is even correct but I am making some mistake in my code, however I can not find it.

from edlib.

Fixed! I added + W to equation. It seems like problem was because query is padded with W cells. I am not sure completely how does this fix things, but it certainly does and it is optimal boundary, there is no better boundary (I tried replacing it with W - 1 and there come errors).
So only thing left to do is examine why exactly that + W helps.

from edlib.