question about IIP3 and P1-dB in mixer

we know IM3 products should rise with a slope of 3 for each increase in the input power,and IIp3≈P1db+10,but some circuits do not behave this way. Because the P1dB of one mixer I designed is about 6 dbm with the frequence of RF 432MHz and the 36MHz IF signal, but when I simulated the IIP3, the result is smaller than 0, and the slope is not smaller than 3. At first I guassed the reason is the sweep range, but after I expand the sweep range, the result did not vary, Now I am confussed with this.
please tell me which kinds of circuits do as this? and the possible reason of this instance ?

first this is a mixer, not a amp ,vottera series don't take account LO signal into.... and the theoratical math approximation is itself a simplified model with respect to amp, you don't know if a1 and a2 a3 changed by the input power level....
at last you maybe exame your mixer for which ismixed with which, we have harmonic mixer just offer the strongest frequency index other than f1-f2, so no one know if it is your carefulless design make your design offer this particular output strong other than what you think as baseband signal p1db

thank u for your answer! i think my design is simple. the figure is inclosed. if the w/l ratio is proper, this feedback can make high linearity, but after simulating, the P1db≈6dbm ,confusedly, the IIP3≈-9dbm.

i can't understand why no one has mad use of this kind of feedback so far? i think this method is better than degeneration using resisitor. also i cant understand why IIP3 is so small.

experts, if u know, could u give me some explanation?