Martin: I don't think you know what you're talking about -- DXO 5.3 does not apply any noise reduction if you turn it off. I'm sorry, but you're just wrong about that. And it's very clear to me that, if you do not apply any NR in 5.3 (i.e., by turning it off, which is very easy to do), then it has a lot more detail than ACR. The demosaicing is sharper/better, and the lens specific sharpening is better.
Look at the second picture (ohne Rauschminderung).
The modul "noise" was definitly switched of.
You can see slightly more detail, than in the third picuture (with NR on: Luminance 25, Chroma 50, the same adjustment I applied on ACR!).
My opinion is, that ACR did - in this case - a better job.
Maybe that 1600 ISO, three stops underexposed on a 5D, was a too extreme experiment.
And I agree with you that in the every day work DxO 5.3 got better results.a
Here a qote from the new dpreview editorial blog.
A comparison between ACR and DPP: "… that DPP applies at least some chroma noise reduction (and sharpening)
even when NR is set to zero which renders the software pretty much less useless for our purposes."
I'm sure, it's the same with DxO.