I agree that there is noting wrong with the approach but let me try to get this across one more time:
there is no difference in the exposure received by the sensor, so describing it as under-exposure is flatly wrong, and misleading.
Thus there is no deviation from "supposed ideal ETTR exposure."
The differences are just in the subsequent positioning of the numeric raw levels, so are essentially just different ways of using discrete numerical levels to encode the signal from the sensor.
Let's compare two situations, leaving DSLRs aside:
1. Camera A where an exposure of 1/60 sec at f8 and ISO 100 results in highlights using the highest possible values in the raw file, just short of clipping, the so called ETTR. Let's assume, by chance, that the level of illumination used in this test is the one defined by the ISO standard,
2. Camera B, say an IQ280, with the exact same exposure results in the raw values being 2 stops short of the brightest possible values in the raw file.
I could be wrong, but according to my understanding of the definition of ISO in digital defined in ISO Standard 12232, sensor sensitivity, rated by its ISO number, is related to the exposure necessary to saturate the camera. Which means to reach the highest possible raw values.
According to this definition, the actual digital ISO of Camera B is lower than 100 when ISO100 is dialed in camera.
For all practical matters, this is similar to what film companies used to do when they rated Velvia at 50 while it was actually an ISO40 film. The purpose was to expose the film less than the camera thought in order to avoid blowing highlights and saturating colors more, in other words under-expose the film.
I fail to see how the approach used by Phaseone is not about under-exposing the sensor per the ISO definition of what ISO rating is.
Now, I believe that the approach used with camera B:
1. Results in more ability to brighten highlights in post processing without running into clipping, which is seen by most photographers as a great thing,
2. Results in shadows having less raw values available per shadow stop, which is not that great for advanced users who would probably want to tap better in the DR potential of their camera, but is still not a disaster.
Now, what part are you not in agreement with?
Cheers,
Bernard