I agree, it doesn't seem like very good design in this respect, Mike. I would have thought there ought to be a couple of resistors to knock the signal level at the CODEC output down 20 dB or more before applying it to the microphone input. I wonder if Yaesu figured this out in the new FT-710?
I have looked to see if there might be transmitted signal-to-noise issues due to this low number of active bits, but setting RPORT GAIN right down to zero and up to 70 does not appear to affect the noise level within the transmit filter passband. So the dominant transmitted noise is apparently generated later in the signal path. The transmit power setting on the FTDX10 does affect the transmitted noise level.
Is there any other test I could do to see if there are other adverse affects?
Despite this issue the FTDX10 does seem to work OK on WSJT-X modes using the USB audio interface, once you set the computer levels low enough that there is no significant clipping (or audio harmonics) and set RPORT GAIN high enough to achieve your desired power output. Whether or not there is an ALC reading (within normal range) appears irrelevant to signal quality - it seems the power control is more like Elecraft's than the typical rig. I've made hundreds (or perhaps thousands - I'm not counting) of FT4/FT8 QSOs and a couple on MSK144 with no apparent issues. The built-in audio and RF spectrum analyzers operate in both receive and transmit, and are very helpful in checking for audio harmonics. But I do want to borrow a good spectrum analyzer and compare what I see on it with what's displayed on the rig's screen.
Some of my notes are at https://www.qsl.net/ve3sma/FTDX10%20Problems%20and%20Solutions.pdf .