Previously I measured the arm power via the power recycling gain and CARM pole. 
Today, using the modulation depth up-down test from November 2022 (alog 66015),
I estimate the arm powers to be 317 kW for 47.1 W incident on PRM.

This is right in between the arm power via PRG (337.7 kW) and arm power via CARM pole (301.0 kW).
The inferred round-trip arm losses from REFL are 64 ppm.
The plot shows the arm power estimates for all three methods.

Details
The total IFO carrier reflection is 7.1%

From the mod depth up-down test results, I found that carrier makes up 90% of the total reflected light and 97% of the input light.

From our HAM1 excursion power budget, I calculated that the total input power to LSC REFL A and B PDs efficiency is 0.00168.
This comes from the following multiplied together:
1) measured old HAM1 REFL to mirror M5 efficiency = M5 refl / REFL beam = 0.0168 (alog 63519)
2) measured new M2 reflectivity = 0.41 (used to be HR) (alog 63625)
3) two 50:50 beamsplitters (M6 and M18) between mirror M5 and LSC refl PDs
4) IFI throughput = 0.977
5) IM4 reflectivity = 0.997573

So given this input to LSC REFL PD efficiency, we can calculate the IFO reflectivity for carrier like:

IFO refl = LSC REFL {A/B} * carrier_in_refl / (input_power * carrier_in_input * input_to_lsc_refl_pd_efficiency)

At the time of the the mod depth up down measurement, 
input_power = 47.15 W
REFL_A = 6.12 mW
REFL_B = 6.05 mW
carrier_in_refl = 0.9
carrier_in_input = 0.972
input_to_lsc_refl_pd_efficiency = 0.00168

yields on average 7.1% between the two PDs.  Differences between REFL A and B likely from small AOI differences causing mirror M18 to not be exactly 50:50, or calibration error.

Code
Code that made this plot is in 
https://git.ligo.org/aligo_commissioning/power_budget/-/blob/master/code/prg_from_carm_pole.py.

Code that processed the mod depth test is in
https://git.ligo.org/aligo_commissioning/labutils/-/tree/master/mod_depth_up_down_test

alogs
June 2022 HAM1 incursion - Day 1 (PSL and REFL path power measurements) alog 63510
List of HAM1 optics, revised alog 63625
HAM1 work on Friday (Ryan1, Craig, Georgia, Keita) alog 63544
Nov 2022 mod depth up down test alog 66407