Before removing #3 of the birefringence measurement setup I measured the incident and transmitted IR power  at normal incidence.

From this measurement, I estimated a rough upper limit of scattered power to be 1 % but note that we had power fluctuations at about 4% during this measurement.

While this value is not precise at all, it is far below the green scattered power (supposedly between 10 and 30 % ; hopefully elog will follow for this measurement)

It means that the scattered power is strongly decreasing with the wavelength.