Following the Gabriele‘s suggestion, to reduce the effect of the power recycling fluctuation, I recomputed the etalon effect normalizing the power transmitted at the end of the arms using B5_DC instead of the power transmitted by the IMC. The result ( ~ usual plot) is reported in the first figure. The black simbols in the curves reported in the first figure have a cleaner behavior thanks to the new normalization and this permit a better evaluation of the input mirror properties. First, must be noted that the B8_DC behavior show a abrupt dropt in the last weeks and this suggested the realignment in the West end bench performed by Julien (see entry 18322).

In the second and third figure the relative power variation for B8_DC and B7_DC are plotted versus the temperature variation of the corresponding input mirror (black symbols). The red curves show the expected behavior according the double cavity model (NB: I found a bug in my previous model and this affected the numerical results... I apologize for that). The parameters used to compute the red curves have been obtained with a (primitive) rms minimization method and they are reported in the attached table. In these figures it is evident that the agreement with the model is good, but some points follow a completely different curve. This occurs for the points after a certain GPS time, as it is shown in the fourth attached picture, where the relative power variation vs time is compared to the etalon model. Around GPS 870307200 the measured transmitted power and the model start to disagree.  This is evident also in the fitht attached figure, where the quadratic errors are computed.

It is important to note (see 6th figure) that the Julien intervention on the West end bench reported the measured transmitted power to the "modeled" value (see very last points close to the horizontal axis), but the power transmitted by the north arm is still far from the expected value. A realignment intervention is probably needed also in the north end bench!