I misaligned the BS and ITMX and looked at the trended time series of REFLAIR_A_RF45_Q_ERR. Since the beam is (in theory) bouncing prompty off the PRM and onto the diode without any kind of Fabry-Perot interference, this signal should tell us something about 45 MHz RFAM.
The calibration (7.6 cts / nm) is from elog 10127.
As Kiwamu pointed out, the calibration must be multiplied by R_PRM / T_PRM^2, since the PD experiences the reflection of PRM rather than the doubly-transmitted beam. The plot is now corrected.