ignoring merge-commits improves the results in two ways:
- on some repositories git-darcs gets caught-up in merge-commits, recording almost only merges (I am testing with 6 repos I ofter work with)
- in the problematic repos ignoring merges increases the amount of recorded patches by a order of magnitude
- in repos with almost linear history ignoring merges has no effect on the amount of patches found
- the history becomes more readable