dsfdr with method='meandiff' vs. method='mannwhitney'

Hello,
In all examples of dsfdr you used method='meandiff', what is the reason for that? when do we use method='mannwhitney'?
When using method='mannwhitney' I always get reject=False even if it seems it was supposed to be True. Is there a bug in the mannwhitney implementation?
Do we need to shift the data so it will have mean=0 and std=1?
Thanks
Anat

Start by reviewing the dsfdr implementations and examples for method='meandiff' and method='mannwhitney'. Reproduce the reported reject=False behavior with a minimal dataset, then determine whether normalization is required and whether the Mann–Whitney implementation is faulty; document the explanation or confirmed bug.

Written by the indexing model from the issue text.