Regarding Metzinger (and similar stuff), it is indeed very astonishing all the time that such nonsense could be published after Peirce, Wittgenstein and Deleuze (I could include Whitehead too:), if he would have resisted mysticism).
Metinger’s argument is weird at several instances, but it immediately fail entirely with respect to two issues: the everlasting nonsense of “states” in the brain/mind, and the more than primitive image of knowledge that he uses in his argument.

I agree that representationalism is probably the biggest problem BNO faces, but it’s gospel in cognitive science for good reasons, not the least of which are the even bigger problems pertaining to the direct realism you seem to be espousing here.

Another thing I would suggest is checking out Andy Clark’s “Whatever Next?” BBS piece on Bayesian predictive coding. By the looks of things, Bayesian models are about to sweep the old paradigms away, and they radically revise the principled distinction between perception and cognition in cognitive science, as well as introduce a new raft of (I think, paralyzing) troubles for the kind of direct realism you’re suggesting. So for instance, the ‘Problem of Illusion’ which so bedevils direct realism accounts becomes, on say Friston’s ‘free energy account,’ an extraordinarily elegant way to explain a number of phenomena, from inattentional blindness to dreaming, that direct realism has great difficulty squaring.

That said, predictive coding does not presuppose representationalism.