We found a mistake in calibration of PIT direction sensing of oplevs.
For PIT sensing, arm length of oplev should not be just the distance between the mirror and the QPD. The incident angle to the mirror should be taken into account. See the attached pdf file for a sketch.
Arm length should not be L, but L*cos\alpha.

Izumi-san told me that this effect is summarized in Sec. 4.11 of FINESSE manual document by A. Freise, so see this document for more detail.

Since I heard this effect for the first time (in fact I "found" it by myself), I guess this effect is not so common in KAGRA,
so I would say it is better to check the calibration of PIT sensing of all the oplevs and to make sure not to make the same mistake in the calibration of oplevs that are not yet calbrated or implemented.