N-path mixer power conversion gain calculation with PSS+PAC

I’m simulating an N-path mixer using ideal components. After performing a PSS+PAC analysis, I observed a positive power conversion gain, while I expected to see a power conversion loss instead. It's worth noting that the beat frequency is set to the LO frequency. To verify this, I performed a two-tone PSS simulation, which indeed showed a power conversion loss. Is there something unusual about the PAC analysis that I might not be aware of? I'm using Virtuoso version 6.1.8-64b and Spectre version 18.1.0.077 64bit -- 1 Aug 2018. Below are the details of my testbench.

I'm using Cadence switches with the following parameters:
- Open voltage: 100 mV
- Close voltage: 900 mV
- Open switch resistance: 1 MΩ
- Closed switch resistance: 1.0 Ω

The switches are hard driven by a square-wave LO swinging between 0 and 1 V whose rise and fall time is 1% of the LO period.

I use the following equations to compute the IF and RF powers and eventually the power gain:
Prf = harmonic((vh('pac "/RF") * conjugate(ih('pac "/PORT0/MINUS"))) '0)
Pif = harmonic((vh('pac "/IF") * conjugate(ih('pac "/PORT1/PLUS"))) '-1)
Gc = Pif / Prf