The current model allows you to mostly interact with the Kindle as you would a book - a light touch anywhere on the visual field (i.e. the page) flips in that direction. However, you can also leverage the added functionality that the Kindle provides, which is indeed the text interaction.
However, I mainly use this for looking up unfamiliar words, rather than for highlighting. I agree that the current system is unsuitable for highly-interactive reading, such as by jotting notes in the margins, highlighting passages, circling words, and so on. On the other hand, I suspect that this type of highly-engaged reading is the exception, not the rule, and that for most of the Kindle's audience, they would no more "interact" with the text of their book than they would if it were printed.
If the primary argument is on the "surprising-ness" of the UI, that could be handled in the same way that Nook did when they first switched to a touchscreen, by overlaying regions on the screen to teach users where to tap. Then, once they were familiar with the behavior, they could turn it off. This is not ideal, as it reduces the legibility of the text, which I'm sure is why this is probably not the default.
Anyway, an engaging read nonetheless, and I'm certainly ready to be proven wrong by a new interaction model that works better. This could definitely be a case of "I want a faster horse" behavior on my part.