Karen,

 Don't think that one tone of the LO or RF inputs is sufficient for simulating
the IIP2 of a stage. There are no simple rules for the number of tones to
use. If you run a transient and plot the LO signal does it look like the LO
signal plotted when from QPSS analysis? To accurately capture the non-
ideality of the LO may require many tones and using 10 tones is not an
unreasonable number of tones to use.

Other questions:
1) Are you relying on device matching to improve the IIP2, that is, canceling
   the even distortion with differential circuits?

  In this case you will need to model the effect of mismatch, that is, perform
  Monte Carlo simulation

2) Have you tried using perturbation analysis to estimate the IIP2?

                                                              Best Regards,

                                                                Sheldon

Hi karen,
             maybe I missed the point of your question, but should you not be thinking about a taylor series description of the signal waveform?

that is: the coefficents for each term f1+f2, f1-f2, 2f1+f2, f1+2f2 etc?

- therefore if your circuit becomes more non-linear under large signal conditions, you should be considering more terms according to how large each term is?

Cheers,
           SimonH.