Suspect land/sea mask from over Antarctica

Hello forum,
I want to mask-select some geolocated data across Antarctica with the following basic commands

region=[-180, 180, -90, -60]
df2 =, resolution=“c”, mask=“s/k”, region=region)
dfbm = pygmt.blockmean(df2, region=region, spacing=spacing)
grd = pygmt.xyz2grd(data=dfbm, region=region, spacing=spacing)
fig.basemap(region=region, projection=“S0/-90/40/20c”, frame=True) # Antarctic
pygmt.makecpt(cmap=“jet”, series=[0, 20], background=“o”)
fig.plot(x=[135, 165, 165, 135], y=[-76.5, -76.5, -75, -75], pen=“2p,gold”, close=True)
fig.colorbar(position=“JCR+v”, frame=[‘x5’, ‘y+l"Kg m^{2}"’], no_bg=False)

However, some strange pie slice appear for both a land (“s/k”) and an ocean (“k/s”) mask.
Now, that region is the Ross Ice Shelf. It could make sense to exclude it from a land mask if land ice is sorted as “wet” surface, but then the same should hold also for the Ronne Ice Shelf before the Weddel Sea and other minor shelves. And this is not the case.
The shorelines (red coded) indicate that the Ross Ice Shelf is land, indeed.
I am puzzled: is it a select bug, is it a GSHHG problem?
What is your opinion?

FYI, I have tested all shoreline resolutions and I report that the pie slice above occurs only with a crude resolution. All those above (from low to full) the maps are just fine.