I had a Motoman robot (welding application) spring a leak in the integrated cooling pipe which caused a massive short that ate through the nozzle collar. I had to replace the entire gooseneck, but we didn't have one that EXACTLY matched the one I removed.

When I cobbled together a replacement, I re-taught the TCP according to the new location (3/4 of an inch stickout) and the robot started rotating around that point. No problem.

BUT, when I went to the TCP test program, it was off the reference point by a sizable margin, and the wire cutting program was also off location (i.e. it totally missed the cutter!).

How is it possible that the robot is still using the old axis positions instead of taking the new TCP to the old points?