Wow, lots of flailing about in here.
So, there are two fundamental issues you guys seem to be talking about here. One is the audio capture not giving you a 'strong enough' peak, and the other is the exported data from FuzzMeasure.
As for the first issue, it's a red herring. You can push that level so hard that you're theoretically "perfectly captured" by the system, and you still won't get "high enough" impulse responses in certain situations. Think about a 'perfect' subwoofer that can reproduce 0Hz to 100Hz, but it cuts off after 100Hz. The resulting impulse response from the deconvolution will not have a peak of 0dB, because there's missing energy from the 100Hz->nyquist range.
So, let's try to address the second one, because that's what matters to you folks most.
When at 16-bit, yes—the output will get truncated once it falls into the noise floor. That problem is "solved" by changing to 24- or 32-bit exports. However, when viewed by other packages, it seems to create trouble because they apparently don't have vertical scale controls? So, there's another way…
The simplest thing to do is to just Normalize the impulse response in FuzzMeasure before it is exported. Just choose Measurement > Normalize from the main menu (or right-click the measurement record and choose 'Normalize'). Once that measurement is normalized, exported signals will also be normalized to 0dB peaks. That way, you can still export at 16-bit with no worries about truncation.
Now, normalizing the measurement will not always do what you want in the graphs. If you're viewing the impulse as an Envelope Time Curve, you can end up with the peak above 0dB. In that case, you should right-click the ETC graph and choose Normalize in the context menu. Alternatively, choose Impulse > Normalize from the main menu.
Hope this helps!