The dictionary of base counts on the reverse strand was not used in the fisher's exact test for strand bias.
For the int(DP) and int(NC) filters, it appeared to me based on their definitions that they were supposed to be compared against min_reads and min_cells.