HTML5test is not a test of standards compliance; the title is misleading. It's a wishlist of features, some of which are standardized, but many of which are not (or are not part of HTML5). For example, html5test doesn't (in general) test whether you've really implemented a feature correctly (or really - at all) it just uses feature detection to check whether you've claimed to implement a feature. Fortunately, browsers are never buggy and this distinction doesn't matter.
Then, html5test follows the whatwg's "living standard" instead of the less-cutting year-old actual standard html. This makes sense at first glance - we want to know which browsers support "new" features too! As a developer, that's great. As a score for a browser, that's questionable. Many features are added to the standard because one of the browsers initially experimented with a non-standard extension; lately that's been webkit/blink due to the mobile push, but previous names have included IE6. By *intent* the whatwg living spec is a few steps ahead of the browsers. What this means is that if you use this as a score is that you're going to penalize whoever is following the spec, and promote those leading the spec. That deserves at least a separate score.
Then, there are HTML5 features that are deprecated, like . The continued support for scores chrome two points, and edge+firefox none. Is that really what you wanted to know? I bet there are *lots* of deprecated features in old IE; if you're going to start counting those...
Then there's features like speech synthesis and recognition. Those aren't part of the spec, have never been part of the spec, yet they're worth 5 points together. Or worse, the Web SQL features, that have explicitly been rejected, also worth 5 points (only webkit-derived browsers support this).
Almost all of the point differences between browsers can be explained by features that are experimental, deprecated or rejected.
In short: don't use html5test. It's pointless.