of course it is the Sony which will blow the highlights at the earliest opportunity if rated normally BECAUSE THE PHASE UNDEREXPOSES BY 1.5 STOPS @ ISO 100 BY DESIGN. ... Overexpose on a Phase back and you have padding, by design, because the ISO is fake.
From persistent falsehoods like this about underexposure and fake ISO values, it seems that some people still do not understand that "ISO" means at least three different things according to the ISO:12232 standard for
Photography -- Digital still cameras -- Determination of exposure index, ISO speed ratings, standard output sensitivity, and recommended exposure index, and in particular that the role of a camera's ISO setting in determining the exposure index is unrelated to any measure of a sensor's sensitivity, and cannot be judged at all from the highlight-handling measurement SSat that DXO misleadingly and somewhat arrogantly calls "true ISO speed".
A) "ISO" refers to several quite different quantities defined in the relevant standard ISO:12232 and people who wish to make accusations against camera makers of fake "ISO" values should at least learn the differences.
1)
Exposure index, a measure of exposure levels as used by light meters, even off-camera ones, and unrelated to the performance of a particular sensor, but used as units of measure for the others.
2) A speed rating for
Sensitivity in the sense of ability to detect light, judged by signal-to-noise ratios and shadow handling. The main standard is S40:1, corresponding to the exposure index at which exposing a uniformly lit subject gives a SNR of 40:1. (Aside: ISO:12232 specifies that usually, this 40:1 standard should be used to describe the sensitivity of a camera -- but I do not see that done very much! For Modern CMOS ILCs, this is probably about 800 or higher, regardless of the "ISO" setting on the camera.)
3) Another speed rating for
Saturation-based speed Ssat: roughly the exposure level (EI) at which metering based on a certain subject luminance level cases highlights to be blown on the sensor (or clipped in subsequent amplification en route to raw files) at three stops brighter.
This is offered as a measure of the lower end of the camera's exposure latitude; a minimum safe exposure index. It is also the basis for what DXO cals "true ISO speed".
4)
Output Sensitivity, relevant only to final JPEG output: level placement in default in-camera JPEGs.
B) The "ISO" dial sets the Exposure Index used by the in-camera light meter (1 above) and controls default JPEG conversion (4 above), and is controlled by the ISO and Japanese CIPA standards for Standard Output Sensitivity and such. ISO:12232 and the related CIPA standard have absolutely nothing to say how numerical levels are used in raw files, leaving that intermediate step entirely to the implementation decisions of camera makers.
C) DXO also does some testing of EI settings, but does not spend much space reporting them, perhaps because the results are rather boring. DXO concludes that on item (1) most cameras are fairly close to the ISO specification for EI: with the same f-stop, same level of uniform test subject lighting, and the same "ISO" setting, cameras will generally choose roughly the same exposure time. (Actually DXO skips the lens and f-stop by shining light on the bare sensor and compares the shutter speeds chosen at the cameras' various ISO settings, in order to calibrate its shutter speed choices in subsequent measurements, and observes that in general, they are close to the ISO's definition of Exposure Index. So no cheating or fakery there!)
So
no camera is underexposing the sensor by 1.5 stops, or giving it 1.5 stops less exposure than some other camera, or 1.5 stops less than some ISO prescription for exposure levels. D) The DXO also measures a version of (2), Saturation-based sensitivity SSat, which in particular is totally unrelated to (1); confusingly, DXO cals this "true ISO speed". As a perverse example of how this measurement could be misleading, changing nothing in a camera except padding raw levels from a 14-bit ADC to 16 bits by adding leading zeros rather than reporting the original 14-bit values will reduce the SSat and "DXO ISO" by a factor of four, but clearly is not reducing exposure level by two stops. This is not entirely hypothetical, given that Phase One does so kind of 14->16 conversion whereas most camera makers do not, so if its values are 1.5 stop further below the top of its 16-bit range than some other camera, the actual numerical levels are higher by about 1/2 stop, so a factor of 1.4 larger.
In summary, all I can see that Phase One is doing giving more highlight-protecting headroom than the ISO recommend minimum, while still using somewhat higher numerical raw levels than most or all cameras due to its greater raw file bit depth.