We have found that the polarization of PDH/WFS path was totally bogus because of one HWP somebody has installed, and that was the reason why the green 50:50 S-pol splitter was giving us something like 20:80. We rotated the HWP such that 50:50 becomes 50:50-ish, but that also means that 80:20 splitter to separate PDH and WFS was also affected.
Before: 4.4mW or so going into PDH, 4.8mW into WFS (0.9mW WFSA and the rest WFSB).
After: 8.2mW for PDH, 1.2mW for WFS (0.6mW WFSA, 0.6mW WFSB).
Tomorrow we'll remove the HWP as it is not in the design.