I drew out the path and also where it could be getting lost. The guide groove is inside the bore which is 4.5mm. Difficult to even see if the contours are well formed by lost wax casting.
Comparing the partial rotations to the CAD model I think it's switching which incorrect path to take randomly. Sometimes it start backwards, others it will start out ok, but get caught on that flat. Maybe it could be adjusted and rebuilt.
