Since the tiltsensor senes the absolute angle of the target, I re-calibrated MCE oplev calibration factor using tiltsensor. MCE P(Y) was excited at 3Hz and 10Hz and the peaks on oplev and tiltsensor were compared. Here are the calibration factors of before and after:

Attached plots are the spectra when MCE was excited in P and Y, respectively. They both worked OK with some factor differences in the lower frequencies; probably they are affected by the air turbulance in a different manner in that frequency region - somehow tiltsensor is better. The strange structures in tiltsensor at lower frequencies (see, this log) are not seen anymore.