It's easy, by comparison, to get to the 80% point with a game -- especially if you have a sample to start from that's somewhat close to what you want to create. What's hard is making every little bit work correctly; creating a finished product that doesn't have glitches, that works on all devices/in all browsers, and that has a good "feel" throughout. That last bit of polish, that last "20%", probably takes well over 80% of the time; I'd estimate it at about 95% of the time. And if you start with an "80%" engine that hasn't gone through that polish phase at least once, then you're doing all that work yourself.
This isn't precisely what you asked: You asked for success stories. A casual browsing of the site doesn't find any published game lists, though. The creators of Phaser.io do have a list of games THEY'VE published with it:
Look at the "HTML5" options; they used to make games in Flash.
I looked at Phaser a while back, and decided against using it. None of those games actually look much more complicated than a proof-of-concept, 80%-at-most demo.
On the other hand, Cocos2d-X has been used in dozens of hit games, and they are up to v3.0 of Cocos2d-js, which looks awesome, has an amazing free toolchain you can use, works in a browser, AND can be embedded in a custom wrapper on mobile (something like PhoneGap/Cordova, only tuned and optimized for game development -- it includes SpiderMonkey directly, for instance, so you don't have to worry about what browser it's running in).
The choice wasn't that hard.
My game was complete, but the garbage collector would randomly crash, every hour or so. I spent nearly two weeks reading through incomprehensible C++ code, and completely failed to fix the problem myself.
Hugh, the creator of NekoNME, graciously helped me after I sent in a deterministically crashing test case, and that's the only reason the game was shipped at all. I consider myself a competent developer, but utterly incapable of debugging someone elses GC.
I also had to make a whole bunch of smaller fixes (eg adding in app purchases, fixing leaks etc), but those things I expected.
If you're evaluating an engine for use, there should either be a flawless game more complex than what you're trying to accomplish, or you should be ready, willing and able to traverse the full stack fixing whatever comes up.
(Note: Haxe is battle tested now.)
But again, it as last year, so I assume 3 years make a huge difference in development, plus Haxeflixel is a haxe framework tailored for making games. To whoever might be interested in diving into it, we shipped a nice amount of polished finished games: http://haxeflixel.com/showcase/games/ (Mine is polaritron)