Dark stripes in coherence map

I am using SNAP for estimating coherence map between two consecutive SLC products using the scheme described here .
Usually it gives good and reasonable results but sometimes the coherence map depicts some dark stripes across the map, these can be wide or narrow but always dark. (see for example the images below) . The stripes discontinuity looks unnatural to me so I assume it’s some kind of an artifact.
Appreciate if someone that encountered this phenomenon or knows what’s its origin is and is there any way to mitigate or disappear it from the map