Contradiction: pnoise+noise separation vs pac or transient

Hi,

I am simulating a circuit with nonlinearity, time variance & frequency translation. The output Vout=(V1) + (V2) .

I used pnoise+noise separation to find transfer function from drain noise of M1 to V1, V2 and Vout i.e. TF1, TF2 and TF and I find that TF=TF1 - TF2 which I want because it cancels M1 noise.

But when I model drain noise of M1 with a current source as shown in figure, and perform pac analysis, I do not see cancellation and I find TF=TF1+TF2

I also modelled it with small sinusoidal current and did transient, performed FFT and found TF=TF1+TF2

I am doubtful which one to trust. Can pac and transient do the job of pnoise OR it is a mistake?
<a title="NTF.jpg" href="../imgqa/eboard/Antenna/rf-4h5r1x31aju.jpg"><img src="../imgqa/eboard/Antenna/rf-4h5r1x31aju.jpg" alt="NTF.jpg" /></a>