New and old scorer return different results when using alignments

[ajasja]

For old sorcerer (v1)

!{CCO_old}/fastscore.exe {fasta} --out-name={v1_bin_bn} --align=-7,0,7 --orientation=P

New version (v3 -- I don't know why I'm calling it v3, but I have this in all my notes:)

!{CCO_old}/fastscore.exe {fasta} --out-name={v1_bin_bn} --align=-7,0,7 --orientation=P

The scores (when not checking alignment) match with the original and the previous version.

More examples at:
03-check-PNIC-alignments.zip

[ajasja]

The v1 seem to return the correct alignments and wrong scores (the P3 P4 pair is incorrectly scored),

while the v3 seems to return correct scores.

More details:
04-check-PNIC-alignments1.zip

[ajasja]

Considering the bug is only in the old solver this bug is actually not that important, although it would be nice to know what was cored incorrectly in the experimental sets...

[ajasja]

The bug is in the old solver, so this is low priority.