As for as I know the IIP3 of the mixer or measure of linerity depends on the gate overdrive of the rf input. The higher the overdrive(Vgs-Vt) the better the IIP3 would be.

In single balanced case u might be using a transconductor as a V-I converter,wheras in double balanced case usualy a tail current source is also used for current matching purposes in left and right side branch. or to say to keep the IF amplitude constant. In that case the allowable overdrive of the RF input must have been reduced. And that could be reason for a lower IIP3. In my view if a tail current source is not used than one can expect the same IIP3 in both cases.

I am a self learner of RFIC design, so may be my analysis is fundamentally wrong. But i wish if somebody can correct me if I am wrong.

Thanks,
--Vikas

yes I am assuming that RF is fed in to V-I converter and LO to switching .