I'm with you on the lack of page numbers... and that would be incredibly easy to fix... insert the page numbers in the actual text body, rather than splicing them in as headers or footers.

OTOH, if e-publishers are OCRing hard copy to create [ kindle | nook | pdf ] versions, they're more hidebound (or living further in the past) than I credit. Talk about do something the hard way!

Modern publishing almost certainly uses a digital version at some point in the process... the author's output; that of a minion who digitized hardcopy and proofed it; a press-(almost)-ready generation, which is to say, 'ready for whatever electronic version of typesetting is in use'... with markup complete. /me cannot conceive a reason for bypassing those, in order to OCR a book.

And even if one imagines a publisher so far sunk in the 'old ways' as to be using linotype technololgy or (heaven forefend) moveable type, somebody's proofing the imprint before it goes to mass production.

Of course, in that last case, someone would have to proof the OCRed version -- before and after [ kindle-ization | nook-i-fication | whatever ] --

                ...which ain't all that damned expensive, given that the typical price of an epub version closely approximates that of buying the text on dead trees!

        oh, was i ranting? sorry.