In this case, it's not really useful to look for the 'main' resonant frequency, and this sort of plot is not very helpful in any case without a lot more information.
'Free' backs, and tops as well, have a lot of different resonant modes. Which one predominates in the response depends on how the plate is held, where it is tapped, and where the microphone is. Even a small change in any of those parameters can result in a totally different set of peaks in the FFT. All of the spetra are 'valid' and give you some information, but it's far from obvious how you sort them out to extract any meaning that can guide you.
A better way to sort resonances, IMO, is by Chladni patterns. You not only get fairly exact frequency information, but the mode shapes as well. My experience suggests that 'free' plate mode shapes tell you more about how the guitar will end up than the pitches.
OTOH, I've been using the Chladni method for a long time, and am still far from convinced that I know just what to do about the back. On one fairly recent guitar that turned out well the back modes that seemed most indicative were the 6th and 8th modes, and were up around 250-350 Hz. The 'fundamental' mode, was down at 44 Hz, and was a twisting mode that tells very little about how the plate will vibrate when the guitar is together. The next one up, at 50 Hz, is a bending mode with more potential utility, but not much more. Either of them could produce a strong peak in a hold-and-tap test, depending on how you do it.