Smaller photosites do tend to have less read noise for given exposure (photon count) received, so closing the gap in the deep shadows relative to a simplistic reckoning based only on photon shot noise, and that graph looks impressive. However, I admit that I do not understand the practical significance of measures of DR at high ISO speeds. It seems that a lot of that can just be the numerous stops of highlight headroom that is almost always irrelevant in practice because no photosites will get that much light. That because—almost by definition—a high ISO exposure is greatly underexposing the sensor relative to full well capacity. I would much rather measure high ISO performance by noise levels, and where the bottom of the "photographically useful levels" is.
I'm very surprised, BJL, that you don't understand the practical significance of measures of DR at high ISO speeds.
Nikon cameras tend to be ISO-invariant, which means if you shoot everything at base ISO, in RAW mode, using the appropriate shutter speed and F/stop for the scene, then the photos that are underexposed due to poor lighting, can be processed in Photoshop to look just as good as the same shot would be at a higher ISO which does not blow any highlights.
The advantage of underexposing at base ISO is that there is no risk of blowing highlights.
However, most Nikon models are not exactly ISO-invariant. There is often a slight DR advantage in raising ISO to 400, say, instead of underexposing at ISO 100. I
n order to understand the degree of this advantage, we need measurements of DR at high ISO speeds.Choosing ISO 400 or 800 instead of underexposing at base ISO, using the same shutter speed and f/stop, might result in a 0.2 EV increase in DR, which is insignificant. However, if the measurements show a 0.5 EV increase in DR, then choosing the higher ISO is significant.
With Canon cameras, there's no issue. Underexposing at base ISO instead of choosing a higher ISO, always results in significantly noisier images.