Hmmmm. Frankly, I'm not sure what the problem is. Unfortunately I don't know very much about the 6356. Are you using the BNC version or the terminal strip version? Because the BNC version has a 4.99 k resistor in parallel with a 0.1 uF capacitor to ground on the AI- input when the switch is in the FS position. That actually shouldn't affect anything, but the gap between "shouldn't" and "doesn't" never fails to surprise me.
The fact that it's differential is no problem. The 4461 is differential, and I believe I usually leave the inputs in differential mode for impedance measurements. The catch is that the system must be grounded somewhere, since the fixture is single-ended. Preferably that's at the generation side, since the 4461 output impedance actually changes according to how it's hooked up (a consequence of the semi-floating differential output stage).
Other things that come to mind are: make sure the sample rates are the same on input and output. Make sure there's not any aliasing going on anywhere.
If you can send me your version of the code, I can try to get my hands on a 6356 and see if I can figure out where it's going off the rails.
Hope this helps,
EBL