Hose the test signal output directly into the, um, I think it's A and B
ports on that instrument, one at a time, and see if you get responses
when they are selected for display. You can just put a filter in-line
between the generator and the receiver channel you're testing and see
at least roughly the response you're getting. Check the receiver gain
settings and display scaling to make sure they are working properly,
assuming you get something reasonable displayed. I'm assuming you can
display A and B independently and not just as a ratio to the
R(eference) input, but I don't recall the details of that instrument
off the top of my head. Clearly it wouldn't be the amplifier that
drives the display y-axis, as you're getting deflection in both
directions with the polar display.

Divide and conquer: is the analyzer OK? If it is, then look into the
T/R test set. If it isn't, look into the analyzer...