However your main point, that GTO can not be exploited, is correct.
People get enamored by theory but forget that the theory depends on subtle details and preconditions that often don't apply in the real world, and end up coming to conclusions that are sound on paper but will not hold up in reality.
You are not guaranteed a draw in poker if you play optimally, you can still lose. It is true that being unexploitable and playing optimally yields a expectation of at least 0 (assuming no rake/fee), but expectation over an infinite number of trials is not the same as guaranteeing a draw.
There is also variance, which can cause sufficient losses in the short run and given that in the real world people only have a finite amount of resources with which to participate, a sufficiently high variance can knock you out of the competition before you ever get a chance to realize the benefit of a long term positive (or even zero) expectation.
In heads up poker, GTO is worst case zero EV. Obviously this does not mean that the worst case in a single hand is a $0 payout - that is a fairly absurd straw man. I can play Federer at tennis for one point and it is possible he will lose. There is nothing interesting in that statement.