I had recently seen samples of the DR the D800 can produce on several websites and it seems to be a fair improvement over the 5D III. Has DxO been giving me the wrong ideas?
The D800 and more generally, newer Sony sensors have delivered big performance improvements at low ISOs. This is only relevant if the performance of the Canon sensor at ISO 100 is not good enough for you.
If you do a lot of your shooting at ISO800 or higher, the difference is largely academic.
In particular, if you don't keep your ISO set to 400 or less (make that 200 if you're using an APS-C body) for fear of degrading the image quality, those differences are largely academic.
I'm not dismissing the DxO results or trying to say that there's anything wrong with them, just that they get blown way out of proportion, and people jump on the numbers too quickly without looking at what the implications are to realistic shooting scenarios.