The percentage of light in the wrong polarization on the two end station ALS tables is measured by dividing the rejected light by the incident light. The incident light is measured with a ~5% beamsplitter. Unfortunately, this beamsplitter is itself polarization dependent, which can lead to apparent percentages above 100%! The attached plot shows these photodetectors as well as the reported percentages. If we assume that the transmitted light doesn't change as function of the selected polarization, the rejected light is in no instance higher than the incident light---even if the reported percentage is well above 100%.

Bottom line: The calculated percentages can only be believed near the correct polarization.