Most of my projects involve a mass-market audience so I get a pretty good view of what average competence looks like. Based on this, I would guess that a significant portion of Americans have a great difficultly reading. Even when you put a big message that says this is not for X, people will continue to do X.
If you run a startup or a company whose audience is early adapters you get a skewed view of the average level of competence of users.
I don't know if things get worse in other countries. However, I would guess that 10-20% of the US population lacks the basic literacy and logic skills to hold a manual job involving anything but repetitive tasks.
And even besides the people with low IQ, most everyone is only capable of thinking abstractly some of the time--and even then only after years of cognitive development. System 2 thinking is taxing to the brain (consumes more glucose/oxygen/etc), and is switched out of whenever it's not absolutely necessary.
It's not about reading ability, it's about the way things are written. There's been plenty of research done on how to write notices and warnings etc. that will catch attention, and how to persuade readers to follow the directions. But of course "user interfaces" are often not "designed" by designers at all, much less anyone who's ever studied the research. (Not that most designers study the research, either, but they're more likely to than Joe Programmer.)
Example: Jakob Nielsen publishes research that shows "people on the web don't read." The sample content used to determine this? The list of tourist attractions in North Dakota.
Garbage in, garbage out.
Finally, the vast majority of text on the web around forms etc. is useless, poorly written, obtuse, abstruse. People have been indirectly trained to ignore it. It's not surprising that most users ignore that messages you took the effort to write.