Blastn's "qcovus" is behaving odd, it seems to be giving artificially low scores

Question

[iddo.nadav]

Hey all,

I'm using "qcovus" in my blast command in order to to retrieve only unique hits between two sequences (no overlap):

blastn -query q.fa -subject s.fa -outfmt "6 std qlen slen qcovs qcovus"

I'm getting a strange behavior..

q.fa and s.fa are single node genomes, with very similar length (<1% difference).

When I blast in one direction (a->b), I get: qcovs=96% and qcovus=96% When I blast the other way around (b->a), I get: qcovs=96% and qcovus=60%

This surprises me, I looked for some bug in the beginning, but everything seems to be working correctly. I dare not suggest Blast has a bug either. Can anyone explain what might be happening here?

Thanks

[GenoMax]

Blast will mask query sequences by default. Can you explicitly turn that off (-dust off) and see what you get?