This work is based on the suggestion in klog32035.

Abstract:

The peaks around 108 Hz and 128 Hz are likely to come from not entire stack structure but the individual optical components.

Detail:

I measured the transfer function from the shaker excitation signals to IMMT1T QPDA2 pitch (left top in fig1), the excitation signals to IMMT2 OpLev pitch (left middle in fig1), and IMMT2 OpLev pitch to IMMT1T QPDA2 pitch (left bottom in fig1).
To compare the TFs from the shaker excitation signals to IMMT1T QPDA2 pitch and to IMMT2 OpLev pitch, I also made a plot that both glaphs are drawn on the same plot (right top and right middle in fig1, red: IMMT1T QPDA2 pitch, blue: IMMT2 OpLev pitch multiplied 2e-3 to adjust the gain to the other TF).

Since white noise injection with entire measurement region cannot make a significant excess on the OpLev signals, I separated the measurmenets into 8 with 4Hz bandwidth (102-106, 106-110, 110-114, 114-118, 118-122, 122-126, 126-130, 130-134 Hz).
Though S/N of OpLev signals are small (~2), the signals can be seen in both signals at almost all frequency bands.

There are clear peaks at 108 and 128 Hz in the TF from IMMT2 OpLev pitch to IMMT1T QPDA2 pitch, which indicates that these peaks are not coming from the entire stack structure but the individual optical component on the IFI stack.
There are also several peaks such as at 118.5Hz but it is difficult to identify the peaks comes from the stack structure duue to the poor S/N of the measurement.