I noticed that the anti-whitening filter for the LF signal of the resonant RFPDs were all consistently inaccurate. So I fixed it.
before: zpk([2.4], [0.24], 1, "n")
after: zpk([2.4], [0.2], 1, "n")
I did this for all the resonant RFPDs, including REFL_A, REFLAIR_A, POP_A, POPAIR_A, ASAIR_A and IMC_REFL. This seems to have fixed the issue of ASAIR_LF going below zero watts.