The 70D...has not improved and is slightly worse than the 9 year old 20D.
DxOMark measures sensors, but people buy cameras, not bare silicon sensors. You can rehash DxOMark data until hell freezes over, it doesn't change the fact that Canon has been outselling Nikon for years, nor the fact that the 5DIII outsells the D800. The obvious conclusion is that 'better' sensors (where 'better' is defined as low ISO DR) have not helped Nikon or Sony sell more cameras.
I would even go further; it is not only a question of salesfigures, it also a question of the qualitiy of the camera. Sure the sensor is an important part, but the DXOmark numbers is somehow like playing topcard with cars, and the Nikon has the biggest engine with the highest torque. But this only shows part of the true real life quality of a car.
Sure, Nikon has some nice sensors at the moment, they are better in some aspects than Canon, and somehow this is reflected in the DXOmark score by a wide margin (the sensors in the Sony cameras are also suposed to be better according to DXO, but just compare them against Canon at high ISO and you realize pretty fast that the Canon sensor gives you better pictures in real life). This score system makes it "easy" for everybody to somehow rank cameras by a single number; this is easy, so everybody does it. But as I mentioned befor, it only shows a small part of the whole thing.
For example, to me it seems that Canon has the better chips, their Digic5+ does a great job when it comes to noise reduction in JPEGs, while the RAWs seem to look a bit noisier from Canon, they look better than the Nikons as JPEG (maybe Nikon shooters are all RAW shooters so JPEG is no priority).
So looking at the camera as a whole, I'm happy with what Canon does and how the pictures I got straight out of the camera look.