b) Why does the amplitude of the output in the transient analysis not
coincide with the magnitude in the Bode plot at 200 khz (452 mV vs. 7.5 mV
or 643 mV)?
You also need to realize that an .AC analysis is a small-signal linearized
analysis. The multiplier, which is an inherently nonlinear device, is
presumably linearized at the operating point, and treated as a linear gain
block.
Thus, if you were to input (say) 100 kHz to both input ports (as you have),
you would not find any 200 kHz on the output ... and LTspice would not plot
the amplitude of the 200 kHz (that this chip actually outputs) ... because
an .AC analysis does not generate sum-and-difference frequencies. If
anything, it will only tell you how much of the 100 kHz goes through
because of leakage/imbalance, and because of the fact that the other port's
bias voltage was not 0.0V.
One might even question if you can do an .AC analysis at all. Whether you
get anything meaningful, depends on what's inside the model for this part.
Depending on how they modeled it, it might work correctly in a .TRANsient
analysis, but not in an .AC analysis. I'm just sayin'.
Regards,
Andy