The current freeze11 db provided for download is problematic, because it contains multiple representative genomes for some specI clusters. This is a result of genome whitelisting in proGenomes.

This will affect metaSNV SNV recall, since unique genome mappings are required as an input. With multiple (too) closely related genomes in the ref db, the number of (uniquely) mapping reads will be erroneously low. Thea @rossum has generated a cleaned version of freeze11 representatives in which she picked the longest genome among candidate representatives.

I'd suggest to update the downloadable db to this version (but see issue #2 re tax ID redundancy).