I think that's often an unwarranted step to take - I'm MUCH more comfortable with the "act as if" formulation than I would be asserting that people actually are even approximately "well-informed" in most contexts. (Or "rational", for that matter.)
To illustrate the problem, consider an industry which is at least somewhat competitive. There is a wide range of available business strategies. Some of these strategies are dumb, but it might not be obvious which. Companies still have to make decisions, so they pick strategies. They might copy a market leader, flip a coin, hire physicists, hire astrologers...but somehow they PICK A BUSINESS STRATEGY.
After a few years, the businesses that happened to pick a particularly DUMB strategy go bankrupt; their businesses collapse and get bought; resources get reallocated to better use.
After many rounds of this game, a LOT of fragile, unlucky, or dumb business strategies have failed. A few remain that seem to work pretty well, and every company still going is trying to follow one of the more-successful strategies.
We now have some idea which strategies work but don't necessarily know WHY. There is institutional wisdom ("we've always done it THIS way") in the winners that they themselves might not understand.
In short, the economic environment (given some simple premises) tends to select for being efficient, regardless of whether people understand why what they are doing is efficient. Figuring it all out does probably happen eventually, but the order of cause and effect might be reversed - you try out the strategy, see that it works, scale it up and perhaps THEN figure out why it works. That's why I can describe companies as acting "as if they were rationally well-informed" even though I know that the individuals involved not only do indeed have a bunch of random "behavioralist" biases (the usual criticism) but in many cases they may actually lack conscious deliberate knowledge of why they do what they do.
The rational actor model underpinning classical economic theory (and much other social science theory) does, in fact, assert exactly this. Of course, the rational actor model is widely-understood to be at best a useful approximation to explain both baseline situations and against which real situations can be contrasted to understand how they should vary from the predictions of the model. (OTOH, challenges to the rational actor model as a baseline in economics and other social sciences argue that its not particularly good even at that.)
> A more accurate premise is that groups of humans tend on average to act at least somewhat as if they were rationally well-informed.
An even more accurate premise is that in some common and important interactions, people tend to be on average be approximately well-informed, and in those interactions, rational choice theory usually seems represents a tolerable approximation of the resulting behavior, suggesting that there are at a minimum important situations in which rationality is a tolerable model for behavior. (Frequently repeated transactions whose utilities and costs are immediate and obvious without obscure, time-delayed, or hard-to-assess utilities or costs resulting from the transaction seem like they should fall into this category, though because utility is not itself directly measurable but is usually inferred from behavior on the assumption that the behavior is rationally motivated, there are some problems with assessing whether this is the case and even whether situations actually fall into this group -- and the inability to measure utility directly risks making rational choice theory something of a just-so story which explains people's behavior as attempting to maximize whatever it is that their behavior actually maximizes. Which ends up not being as much of an explanation as a tautology.)