I'm not sure what I'm looking at there, but I'm guessing that is the mill drawing with a pen in the spindle.
Circular (good) interpolation requires very low backlash. The trouble usually shows up at the quadrants. Where X or Y is coming to a stop and then reversing direction. Any backlash means the transition to the reverse direction does not occur on time and deforms the circle. That seems like it could be what we are seeing there. X continues on moving but Y has not actually started to produce the reversed motion yet even though the motor is turning the screw correctly. When it finally does 'catch' and start moving, that creates the weird flat - to - elbow mark in the circle.
Check your Y axis backlash all around the table. I suspect that is the (main) culprit at the moment. Recheck all the mounting links from the ACME nut to the saddle on down to the bearings in the Y axis motor mount. Looks to me like something has some serious slop in it.