In true it deletes not 89 files, only 48 (from 89 selected files, only 48 uniques, others - repeats from diff. group). For speedup we count only selected rows not checking uniqueness of file (one file can be in a few groups). In future we possible change this behaviour.
I think I understand you issue, it our small mistake, when scan continues, counter of new files begin from 0. But it work correct all files scanned before not scanned again they just not counted. In next version we fix this, counter will begin from previous value.