CONF, I can appreciate that there would not be any trim-follow-up (as in manual flight) whist the AP was engaged. However from page 1 of the report, with the AP engaged (9 29 50) back stick appeared to result in nose down THS, possibly due to an autopilot generated trim demand to balance the opposing stick input.
The AP was still attempting to follow the ILS, but the mechanism of control had to balance the computed demand for the flight path with that of the disturbing demand of nose up stick input. Thus THS movement opposed the stick input and the AP managed a semblance of GS control.
Following the AP disengagement the aircraft was out-of-trim nose down, thus a tendency for nose down pitch until both pilots applied back stick. Presumably thereafter in manual flight there was some manual trim-follow-up.
There is still no explanation as to why the an opposing stick input (fighting the AP) at such a low altitude did not result in AP disengagement or provide a further alert in order to avoid a hazardous out-of-trim condition.
In systems where the AP design requires a higher force before disengaging then a combination of force-duration may be used, but again in this situation with a low force, the duration appeared unacceptably long for the low altitude.