It's unclear how it handles event listening, but even so, they should fork React and replace the rendering target with their library, then use that to benchmark. It'd just be for the sake of benchmarking (e.g. not something they should maintain or encourage people to use).

It's one thing to say "this is better than virtual DOM". It's another thing to show it with numbers across common browsers.