But I don't see this as an issue because the measurements are all equal for all cameras. So if a camera is ranked at 12 stops, and another at 10, I certainly would expect a comparative difference here.
Only if the SNR curves are linear, which they are not. That was my point. For the GX1 and GH2 for example going from 0dB to around 12dB takes the difference in DR from 0.9EV to 0.6EV. Not a huge difference, to be sure, but not the same either.
This is also why the MF sensors look so "poor" on DxO. Using 0dB as the reference the K5 vs the 645D gives you 13.6 vs 11.3. Using 20dB gives 8.4 vs 7.5. That's 2.3 stops difference with one measure, 0.9 stops with another. The choice of DR measurement level can't be arbitrary without imparting an arbitrary bias to the
differences between sensors.
It is just the problem with trying to take a complex system and boil it down to a single number. Certainly the numbers do indicate which sensor has the lower read noise, and that means the lower shadow noise as well. The issue is that the differences can be arbitrary magnified. They of course don't include things like WB which can be as large a difference as well - especially when you consider most low light shooting is tungsten.
So when you see differences at a fraction of a EV on DxO one should be really careful to start looking at real images instead. When you see huge differences (e.g. K5/D7000 vs m43) then you can be fairly confident you'll see a noticeable difference in real world shooting.
So I still don't understand how DXo doing it that way still fails to account for real world photographic DR we can use.
It's just a question of what SNR level to use. Why not use -20dB instead? Then the GH2 and K5 advantage would be even more EV. Why 0db? Because it is a round number? Go push shadows to the level that DxO claims for the dynamic range and see if you consider them useful - they aren't. So maybe 10dB or 15dB is a better number?
Again, if you could arbitrarily chose an SNR number and it wouldn't affect the differences between cameras then that would be fine. But that is not the case, what level you choose
does affect the differences. Not necessarily between all cameras, and not necessarily by huge amounts, but on the other hand you have forum nuts belittling sensors that have DxO DR measurements of a fraction of an EV lower than other sensors and as just demonstrated the choice in SNR level can affect the measurements by that much or more.
Ken