Long time I havn't flown a "N"
but, I remember we also can check the full needle deflection on each galvanometer when we select the test switch to "OFF" just before disengaging AP. With only half a deflection, we can expect a lazy and slow correction on affected axis. It also worth having a look on the coupled axis FD bar to check if AP correctly responds to required corrections.
Often the fault is originated by actuators amplifiers....but don't take it for granted, to me it's more like esothery and magic stuff.