I would agree with Jack's note of the exacerbating issue but disagree with his proposed solution.
The thing than makes most SE sites so much more useful than forums on the same topics is that not everything is displayed equally. Forum threads with posts arranged by date tend to fill up with a lot of dross and finding the actual (best) answer to a question might involve quite a bit of scanning. Frequently this involves trying to figure out what posts to ignore, what isn't even about the same topic etc. On SE sites the voting system creates a crowd sourced filtering mechanism whereby the most useful posts rise to the top and inaccurate or less useful ones sink.
My observation of the "problem" post is that the site is actually working as it should. The dubious post that does not represent sound linguistic research has been heavily down-voted (before even you made a stink about it) and a competing answer with more widely attested basis has been up-voted. From this I conclude:
At least a decent percent of our user base is savvy enough to recognize the difference between sound and dubious linguistics and votes appropriately. As long as some majority of voters are up-voting quality content and down-voting bad, this site will continue to be viable.
The site is not a "crap shoot" in the way alleged. In my opinion people shooting from the hip tend to have a lot more time on their hands and hence tend to be more prolific contributors than true experts. Our true experts tend to have higher demands on their time elsewhere. That being said good answers do come along eventually, and they are pretty well received when they do.
I don't think deleting incorrect answers is a good plan. In fact I think it would be more useful to the site overall if such answers actually stayed and the heavy down-voting would show that consensus believed them to be wrong. I don't see any problem with having cooky personal theories in the answer mix as long as the ones that don't stand up to scrutiny get sufficiently down-voted to differentiate them from more grounded material.
The problem in this case does come back to the sort order being overridden by the OP's choice to accept the answer. I have long since agreed that down-voted answers should be exempted from that display behavior just to avoid cases like this, but honestly I don't think it's a big deal. The important thing is that such posts get extra down-vote attention.
While I don't agree about the alleged problem, the solution I propose if there were to be some validity to it would be:
Vote up good content, vote down bad content.
Do research and post good better answers anywhere only dubious ones are in play.
P.S. Offsetting the fact that accepted answers are kept at the top, answer posts down-voted to -3 or more are shown as greyed out. This is a good interface paradigm that draws strong attention to the fact that the crowd-source mechanism thinks you shouldn't rely on the content.