Thanks for doing all those measurements, if my understanding of U225 is correct (the manual doesn't describe its internals) it is something called a 'Gilbert Cell', which functions as a current multiplier. So if one current input is steady and the other is the signal then the steady current controls the gain of the cell for the signal input. You can also change the polarity of the steady current and make the amplifier effectively become an inverting amplifier. Being current driven, its inputs are always close to zero volts and I am not sure how significant are the small differences that you see. If you can find R551 on schematic <5> you can try adjusting it to get the +9V back to its nominal value and see how much the calibration changes. The -9V is derived from the +9V via an inverting op-amp and has its own adjustment, R570. Adjust +9V before testing the -9V.
In order to get the 7A12 working more nearly to its spec you can use the front panel gain adjustments to try and get the two channels somewhere near their nominal value. If you can borrow a second vertical plugin that would help enormously in being confident that the errors are in the 7A12 and not the mainframe.
Any progress in removing the 7A12 from the mainframe?