In order to exrapolate the round trip losses, I have looked at the change in the reflected power during the "meta-unlocks", where the cavity is unlocked but still resonant. Unfortunately most of the PDs do not see a very clear signal.

In particular, as already observed sometimes in the past, B4 seems to get more light when the cavity is locked. This is an unexpected behaviour which has not been understood so far.

For what concerns the west arm, I'm not able to extract any loss number (see pic 1), while for the north arm (pic 2)  the change in the reflection is recorded quite clearly by the two B5 photodiodes.

They give losses of 390 ppm (B5_PD1) and 250 ppm (B5_PD2). I remark that B5_PD1 has a large offset of about -0.445 mW while B5_PD2 has an offset of -0.0093 mW.  

B1p_PD1 also sees a signal, but the reflected power during the lock keeps growing and probably it cannot be trusted. Anyway, it corresponds to a much lower loss level of about 20 ppm. The offset (measured during a IMC unlock) is about -0.2e-3 mW.