A regression calibration method for rotating element ellipsometers developed by Jobs [B. Johs, Thin Solid Films 234 (1993) 395] was applied to calibrate a homemade rotating analyzer type ellipsometer (RAE). The calibration results further confirmed that the Johs calibration method is valid, even if the responses of the optical elements are not optimal. However, it was also shown that the Johs calibration method may work better on optimal samples, and it may work poorly for some samples if the systematic errors for the ellipsometer system are too large. This conclusion does not contradict Joh results, but extends them by presenting caveats of the Johs calibration method.