This is often a result of designers failing to step outside their "designer window", if you will, and step into the users' shoes. It's possible to make something look right and translate that into conversions if you have enough of an outside perspective.
That said, it's pretty easy to get too involved in your own work to notice a bigger picture so breaks to step back are highly recommended.
When I was a/b testing UI designs for a stable of eCommerce sites, quite often the blatantly fugly design would convert much better. Especially for low-price-driven sites. My personal theory was that people figure the unpolished, crappy looking site is more bargain-driven (kinda like a dingy dollar store in real life).
Sticking a giant firetruck-red "CONVERT" button surrounded by flames and rainbows really does sell better. But boy does it look awful.
This is also evidenced by Amazon.com still looking really crowded and ugly. They test every last thing and if this is what they're using, it's because it works.
This same concept goes for those horrendously ugly, cookie-cutter, inline sales pages plastered with maroon Comic Sans headlines that have highlighter-yellow backgrounds. I've been there and done that, and the facts prevail. As a good/(as in one that creates aesthetically pleasing things) designer, you actually suck conversion wise.
I haven't done much research behind it, but I believe this has to do with why the flat UI is so pleasing. I can only imagine that it is easy for anyone to notice that it's beautiful regardless of technological/design experience/knowledge.
I am glad you pointed this out, your premise is a quintessential example of "UX is not UI."
It's not the visual or web designers primary job to A/B test or step into the users shoes on every detail, that's what a responsible and valuable UX person should be doing (note I didn't say UX designer).
That said, it's pretty easy to get too involved in your own work to notice a bigger picture so breaks to step back are highly recommended.