Non-matching sequences achieving much high scores than exact matching sequences

[hmusta]

I've noticed that in some cases, sequences that have no significant alignment to a reference achieve higher scores (and much lower e-values) than those that align perfectly.

In particular, this occurs with the NC_000913.3 reference sequence when aligning the following reads in a file called test.fa

>NZ_CP011331.1-92/2
GGCACGACTGATTCCGCCGACTCCTGTGTCCACTGCACAAAGTCCTGTTGCAGACGGTCACGGTTAATGCCGGTCAACAGCCGGGCGGCAGCAGGCGGTATATAACGCAGCGGCGAATGCAGCAGGGCCAGCAGCAACACGCCGCCGCGC
>NZ_CP031912.1-287/1
ACACGCCGGACTGACTCTGGCGGGTTGGGTGGCGAACGATGTTACGCCTCCGGGAAAACGTCACGCTGAATATATGACCACGCTCACCCGCATGATTCCCGCGCCGCTGCTGGGAGAGATCCCCTGGCTTGCAGAAAATCCAGAAAATGC

The first read has very few matches to the reference, but BlastFrost reports and e-value of 1e-427, with the underlying score being 334, whereas the second read matches exactly, but gets and e-value of 1e-427 and score of 136.

I built the graph using the command

./Bifrost build -k 15 -r NC_000913.3.fa -c -o NC_000913.3.bifrost

and aligned with

./BlastFrost -g NC_000913.3.bifrost.gfa -f NC_000913.3.bifrost.bfg_colors -q test.fa -k 15

I am using commit d425ac6 (BlastFrost) and 82adeedd5eb1007fdfae2324b339ae32c1c128d3 (bifrost).

Any guidance would be greatly appreciated.

Best regards,
Harun