The n_peak_inconsistencies metric still has a good amount of false positives. A better parametrisation is needed. This might also be an issue with a too low number of cells in the testing resource See #31 for more info
