1) The set of functionality it tests is pretty arbitrary. Some of the things it tests are basically single-browser-vendor with other browsers having clearly said they have no plans to implement them.
2) The point values it assigns to the things it tests are totally arbitrary. So you can implement various useful things but have fewer points than one less useful thing.
3) Its tests are easily gameable (and commonly gamed by various mobile browsers, though less often by IE/Chrome/Firefox/Safari).
4) Some of its tests are just buggy. E.g. it claims that a Firefox nightly doesn't support "script execution events", whereas in fact Firefox supports them just fine. It's just that the way the test runs its test script is relying on a browser bug that got fixed, so the script doesn't run at all.
So I would take any claims based on its summary score as pretty much meaningless. Results for specific things one cares about can still be meaningful, as long as you keep #3 and #4 in mind.
However, I would never take anything you say, Boris, with a grain of salt. Especially since, about 8 years ago, you tried to recruit me to work on Firefox ;)
It's just that people like to forget all this and treat the numbers they get as revelation from on high. Sort of cargo-cult quality measurement.
The list of supported tech is getting pretty real as well [1]. Not everything yet (wot, still no desktop notifications? do you even want people to use your browser for real work?), but a lot - e.g. very much of ES6 is in there and usable right now from the devtools console.
[1] http://dev.modern.ie/platform/status/?filter=c3e0000bf