I'm not a big fan of this test. Since each camera might handle differently, I'd want to identical framing and optimal exposure for some bright element of each, then we'd look at the shadows. My question is not how each looks at the same exposure, but which scenes can/can't be captured with reasonable use of each. If one handles highlights better, why is it wrong to increase exposure to make use of that? Setting equal exposure doesn't seem like a valid test to me.
You don't think how they look at the same exposure is a valid test? Wow.
If one has more highlight range then a "best possible" test would exploit that. That said, I'm not sure to what degree this is the case if at all. Just pointing out that it is something to consider.
Expose to the right, with less dynamic range to start with? Why should you even have to?
The dynamic range in your test is nearly identical. The exposure latitude is what's different. And ETTR is for every sensor, Exmor included. Having less read noise and therefore better shadows doesn't eliminate the fact that the last few bits have almost no tonal separation if you push them hard enough, an inherent fact of linear ADCs. With digital you want your exposure to the right without clipping highlights if you are going to maximize DR and latitude in post. (If you're not then none of this matters.)
When a Camera is not able to differentiate noise from detail in the lower 3rd of the tonal range, then no test is going to make it shine.
This is not an accurate evaluation or statement. If the noise were that bad you wouldn't have been able to push the Canon RAW 3 stops at all.