I have noticed bugs in the solid boolean operation when I tried to unify two Breps extruded from a surface to opposite directions.
It seems there is orientation sensitivity not apparent in the Rhino function.
It might be a known problem, but I will attach a simple ghx to show this problem just in case.