Everybody talks like Sony/Nikon have some sort of amazing technological advantage over Canon. Except for DXOMark scores there is nothing that shows this. Dispite what DXOMark says tests at DPReview show that the 5D Mk III has more dynamic range than the D800 http://www.dpreview.com/reviews/nikon-d800-d800e/19
So either Nikon has some really crappy processors that mess up the 2.7 stop advantage that DXOMark says the D800 has, or we should stop quoting, worrying about, or looking at what DXOMark says.
Or dpreview is using JPEGs and not RAW data.
While you're on that dpreview page, make these changes:
Set the "Canon EOS 5D Mark III" to "HTP On"
Set the "Nikon D800" to "ADL Extra High"
This should give you:
Canon EOS 5D Mark II - ~33 to ~2
Nikon D800 - ~38 to ~2
The whole point about DXOMark is that they are supposed to be measuring the sensor without all the electronic nannies.
So, yes ADL and HTP can stretch the data to expand the dynamic range, but, DXOMarks claim that the D800s sensor alone has more dynamic range just doesn't hold water.
I think perhaps that you are in the "a little bit of knowledge is dangerous" basket.
If the D800 sensor wasn't able to record such a large dynamic range then where would the data come from for the output above where the D800 can reach down to 38? That number isn't achieved with some magic. It is just a different tone curve applied to the raw data for the rendering of the JPEG output. The tone curve is the means by which raw values from the camera's sensor are mapped into the JPEG colour space.
To put it briefly, there are 16,384 possible values for each of red, green and blue in the raw data of current model DSLRs. For JPEG files and your monitor, the computer has to squeeze them into a value between 0 and 255. What the above graphs mean is that the Nikon sensor is able to generate meaningful input across a larger part of that 0-16383 number space than the Canon 5D Mark III can but for a normal JPEG output, a large chunk of the dark tones are simply chopped off as being black. Now with the "active lighting", you can change the way the numbers are mapped and move the black point from (say) 2048 on that 0-16383 scale down to 10 or so. With the Canon sensor, that's not possible as there is no meaningful output from the sensor below a value of 2048.
Now it may be that you still don't grasp that. Ok. Go and read some of the reviews of the D800 and look at what people are able to do with shadows - they can get meaningful detail out of the shadows without introducing any amount of significant noise. You simply cannot do that with any currently available Canon DSLR.
DxO are right and their measurements have been backed up by numerous folks with D800s when working with raw files in Lightroom, ACR, etc. dpreview's results match up with DxO's, you're just misunderstanding all of the data that dpreview are giving you.