The triode mos degenerated pair, will probably exhibit a 3rd harmonic signature, which is sometimes slope not equal to 3, depending on the input tone strength. This is because the triode mos will not be 3k resistance fixed as is modulated by input signal strength. Try plotting the current through the triode mos vs the voltage across it, it won't be 3k fixed like a real resistor.
You could also see the difference by purely doing an input dc sweep, and taking the differential of the (differential) output current (which is gm) and plotting versus vin. For small vin the gm's will be the same between real R and triode R (if =3k). As the vin increases, the gm characteristics will deviate between the two graphs, showing Rtriode shift from 3k. If the slope of gm vs vin for the triode degenerated pair equals that of the real r case, then your IIP3 will grow at 3db/1db, otherwise it will deviate, as yo have seen.
hope this helps.

the tone strengths of 10uV to 200uV already seems very small. The 2*f1-f2 distortion output must be very small indeed, i wonder if you are suffering simulation accuracy issues/can you easily see the distortion tone above the numerical noise of the dft ?
Why are you using such small input tones ? If triode degeneration is indeed a fixed 3k in the transient run, then there doesn;t seem to be a reason why it should differ to the real r case, so try what i suggested in the previous post.
Have you read the IIP3 paper under the analysis tab of the designers-guide.org ?