^ Yeah. As I recall, the Avisoft noise tests are made with the recorders set to their maximum input gain. Noise performance with input gain set considerably lower such as we tend to use them can be quite different.
My own anecdotal experience with DR2d is that even in live music performance recording situations where the ambient noise floor is the lowest I encounter by far (a couple purpose-built acoustically isolated classical music performance halls), the ambient noise-floor is higher than the noise of the recording chain itself including the DR2d, so as long as the gain-structure is setup correctly the ambient noise is what dominates the nose-floor of the recording. That's with mic-in and line-in on the DR2d set to the equivalent of unity gain, using sensitive mics (20mv/PA, 100dB S/N), optimally powered by a preamp providing the gain required (minimal, dialed down to unity or even slightly below that for more highly amplified material).
Good gain structure is a big part if it. I assume the noise-floor of the recording chain itself is determined by the self-noise of the microphones I'm using, since they're not exactly the quietest available (A-weighted = 23 dB Typ. (re. 20 µPa); ITU-R BS.468-4 = 35 dB Typ.) but apparently that's still slightly lower than the ambient noise-floor in those halls.
[edit to add that I've not used A10 in that recording chain, so am unable to make a direct comparison]
