Simon

On Friday and Saturday, measurements with equally mixed S and P polarization in the input beam and pure P-polarization have been finished. The resulting maps and the statistical analysis can be found in the attached figures (45 deg refers to mixed S and P, and 90 deg to pure P-polarization).

As can be seen together with the results of pure S-polarization, the out-going beam is never reaching the input beam's polarization. We are off by at least 11 degrees (15 deg in average) polarization angle, which is a strong hint that the out-going beam has become elliptical. The average for the linear incoming and out-going beam is apparently shifted by 2-3 degrees. This is obvious for the mixed S and P polarization as the out-going beam clearly has its average at 47~48 deg, while the incoming beam has 45 degrees.
I can confirm, after doing some tests by rotating the incoming beam, that the maximum S and P polarization are each shifted by ~6 degrees compared to the incoming beam when the test-mass is in place.

Due to the elliptical out-going beam, the angle-distributions at both pure S and P polarization have half the standard deviation as the mixed polarization due to the fact that we see the change in the polarization angle to both sides in the mixed input-beam case, while for the pure cases this view is limited by the trigonometric functions which we are using here for the analysis.

Apart from the average, which shows the behavior as discussed above, there is a prominent region that is obviously different (top-left). Here the input beam experiences probably a rotation by ~16 degrees in addition to become elliptical. That would explain why the plarization angle changes only little from S to mixed S and P polarization.