Not sure how clustering exactly works, but it doesn't seem to take complete different artists into account. Preferably it should calculate the distance between two artist names of multiple files, and only cluster/group files when the artist names have only a small distance between them (i.e. match Madonna and Madona but don't match Mariah Carey).

Background from https://community.metabrainz.org/t/feature-suggestions-after-matching-a-huge-messed-up-folder/712286: Sometimes (often?) clustering does not differentiate between artists, but a lookup still matches a release so you end up with a release that has too many files. But due to the default columns in the matched release view, it is not always immediately clear that this is because of completely different artists, I had this a couple times with Greatest Hits albums where I was trying to find a matching release but only after unable to find one I realized that it was just because clustering had combines files form Artist A & B