Let’s not mix things up.
Masking before unwrapping helps to prevent unwrapping erros in areas of phase noise.
Masking after unwrapping (e.g. of the final results) can help to prevent false interpretations of mathematically derived results (e.g. unrealistic displacement)
They serve different purpose and depending on when you apply it, you might get different results. In the end it depends on the spatial distribution and proportion of low coherence areas.