The issue is due to the fact that Power BI does not always know which side of a many-to-many relationship to filter by.

In SQL -- joins produce a "cartesian product", or, a raw join like you would in a "merge" operation in Power BI. In Power BI, relationships just work via filter propagation to determine which rows to match in certain situations.

If you need to avoid this kind of ambiguity or it's causing problems, try merging your queries to bring in the data fields you need via a join, that's what I typically do when I need to avoid many to many relationships.