To become a good game designer, one has to design games and-- here's the important part-- get feedback from human players. That seems tautological and obvious, but it's an important insight. What seems like a great game will be broken in ways a designer might not anticipate once it meets real players. Non-designers might find the brokenness of the earliest Magic cards to be laughable-- how could they not see that Ancestral Recall was broken?-- but designers get it. Balance is hard, man.
It's hard as hell to come up with a game worth playing. I probably designed a hundred crappy ones as a teenager before I got a hang of it and, even then, I only have one real success which is a card game-- a fairly closed system, not something continually expansive like Ultima Online or Magic: the Gathering. It takes a lot of trial and error to get good at it. It's easier to go through that process when you have access to other nerdy teenagers who'll be playtesters. For adults, it's much harder to get testers. That's a huge limiting factor.
Then, combine this with the big-budget corporate setting. Few game designers get to design a whole game. They work on pieces, not having responsibility for a whole. It's the same issue that we have in software engineering: most programmers never get good because so few ever work on real stuff in their paid gigs. I haven't worked in AAA game development, but it wouldn't surprise me to see the same principle in force: mediocre designers because most corporate work is mediocre and people never get the chance to get good.
I think Garriott has a point, though, in the assessment that it's rare for a "pure idea guy" to be anything but mediocre and passive. Curiosity is an important trait in a game designer. If you've got it, you'll code. That doesn't mean that you'll do it professionally or be able to write x86 assembly by hand, but you've no excuse not to fire up Codecademy and learn the basics of programming. Most of the great designers had other interests. You can see Garfield's CS and math backgrounds in the design of Magic. You can see Paul Graham's painting experience in his approach to programming and language design. The biggest hits require crossing wires.