On keita's (and daniel and stefan's) suggestion, I went back into the PSL to check on the ref cav power budget.  

First I measured the AOM diffraction efficiency, measuring with the filter:

Before the AOM there are 34.5mW, after single pass there is 19.6mW, after the second pass there is 14.05mW,  40% double pass efficency.  This has degraded since christina's meassurement of 72% in Link  I measured the RF power to the AOM by first disconecting the input to D1201423 (VCO AOM driver in PSL racks), then connecting the cable to the AOM (after the blue SLP-100) to an RF analzyer, then reconnecting the input to D1201423.  there is 30dBm of 79Mhz.  According to Christina's alog it should be 1.2W, I'm not sure if the 0.2W matters.  Since the cable was attached to the AOM verry tightly and the translation stage moves, I could have disturbed the AOM alingment in the process.  

Measuring without the filter, there were 11.9mW incident on the ref cav (after the circulator), and 6.9mW in transmission.  I measured 11.3mW after the circulator when the REFL PD DC ( measured using the lemo cable connected to the TNC DC output) had 66.4mV above the dark offset of 0.8mV.  When the cavity is locked, there are 14mV above the offset on this PD, meaning we have 2.4mW of power in reflection.  This means we are loosing 18% of our power somewhere.  We are probably loosing a few percent in the circulator, but that is at most 5%.