Quick testing on the same curve & surface:
So now, you can just hand over the code to the PointDeviation devs and have them translate it to C/C++ to replace their shoddy code with! 
EDIT: Similar to this thread, I now did a Pull with Loose=Yes and got a way, way worse result. A result outside of my tolerance even. I guess it perhaps should be expected in this case, but I’m still so annoyed that Rhino never warns me when a tool made a result that’s out of tolerance! (But hey, I guess that’s one aspect where Rhino is very consistent…)