using the coherencethreshold of 0.2 in the Goldstein filtering will not automatically set the phase to zero at these pixels.
The masking you mentioned is effective, we have also discussed it here: Filter out low coherence pixels before phase unwrapping
These areas are treated as equal in the unwrapping, but they will not remain zero, because unwrapping adds up adjacent phase values to an absolute surface. The only benefit from masking before unwrapping is that these areas will not introduce ramps in your unwrapped data because random phase is integrated.