A person who close-mikes The Who (a band I heard live in 1967 or 68 and my ears are still ringing) with high-output condenser microphones might never hear the inherent noise of a given recorder since it will always be masked by signal, while the same recorder might be considered unusable by someone who records clavichords with ribbon microphones placed 20 feet away (although granted, that's a dumb way to record a clavichord).
All I'm saying is (a) give peace a chance and (b) the range of signal levels that come out of all our different microphones under different circumstances is so wide that any noise measurement of a recorder can only be meaningful for people who use the same or similar gain settings as the one used in making the measurement.
An instructive example: I own eight or ten different preamps, and the one that measures the quietest at very high gain levels is NOT the quietest one at the gain levels I actually use when recording--and vice versa. The one that's the quietest at the gain levels I actually use isn't a preamp with a very good reputation for being quiet--because that specification is measured at the preamp's maximum gain. Yet it outperforms all the others at realistic (for me) gain settings.
When I measure preamps and recorders, I use my own typical recording assignments as the basis for deciding what signal levels to use for testing. The noise, THD+N and overload limits that I measure are most useful to a person whose requirements are similar, while if your requirements are markedly different, my results may not mean very much to you. That's just how it is.
If I wanted to do everybody else's homework as well as my own, I might test at three different levels (say) 12-15 dB apart from each other, but it's still essential to have a constant reference level when comparing different pieces of equipment. The way of finding a nominal signal level for testing that mshilarious describes has the virtue of greater fairness to different recorders that are designed for different purposes and usage scenarios. Each one gets evaluated on its own terms. But I'm just an egocentric bastard with a test bench and limited patience, who's trying to find the best options for myself and for others who record in fairly similar ways. I'm not normally as concerned with the details of a recorder or preamp's performance if it's unsuitable for the kind of recording I usually do.
Basically I know what signal levels my main microphones usually put out during a recording, so I measure preamps and recorders at that signal level. In fact I use my microphone bodies as part of the test setup so that the source impedance driving the mike inputs will be realistic (you can substitute a fixed capacitance for the capsule and push audio signals through that capacitance into the body of the microphone--a nice way to avoid the non-repeatability of "shootouts"). If a preamp or recorder can't handle the levels, that tells me all I need to know about that preamp or recorder; I don't normally readjust the testing levels to accommodate a preamp or recorder unless I'm trying to document the clipping point of its inputs ("how bad is it, exactly?").
--best regards