I found the error in my calculation which was actually not really an error but a non sufficient precision of the calculation software. I made the calculations again using Mathematica with higher precision.
The results can be seen in the attached graphs. I uploaded the results for each mirror and type of vibration (horizontal/vertical).
As Akutsu-san pointed me to this, I should add that the strange curve of PRM(horizontal) is maybe due to the high peak of its transfered vibration noise at around 0.5Hz.
It seems that I first have to dump this peak to become a more realistic result...