If you take a look at slide 15 of https://www.modalshop.com/filelibrary/Excitation_Techniques_IMAC_30_by_Pete_Avitable.pdf, you will see that a good impact excitation force is approximately equal across frequency bins. Your plot of Gxx has a dip at sample ~10 which we concluded is 10 Hz. Your response has a peak at about the same frequency. From this I would conclude that the excitation force was not good, and that there is a resonance at ~10 Hz. If your frequency spacing of the response and force signals are different, which they will be due to the different sampling rates, you cannot obtain valid cross spectra, so just view the response spectra.
The Raspberry PI can handle up to 8 of these HATS. Its bus limits throughput, so by recollection it can handle 3 channels at 51.2 kHz, and the maximum sampling rate decreases with the number of channels. For modal you are not likely to need these high throughputs, perhaps limiting for envelop analysis of bearings at higher channel counts.
I do a lot of vibration analysis, including modal from time to time.