problem going across Greenwich with PlotMostLikelyQuantiles

I discovered a bug with this function. If we plot the likely terciles over Europe, it's usually good:

But if I plot happen to go from a negative longitude to a positive longitude using a mask, suddenly, a line appears:

But also, the mask doesn't seem to work on the western side of longitude 0. I used the RPSS to create a mask:

and the data with longitude >0 are masked correctly, but the data with longitude <0 are not masked at all.

Thoughts @nperez ?