Public domain books scanned by volunteers are likely to have some errors. You could always help out by proofreading as you read the books, if you want. Myself, if there aren't too many errors, I overlook them.
I wonder if OCR will improve, some don't work well at all. I've been working on converting a PDF scan of an old book, but it has been slow going. The OCR program I use (I use a free web-based OCR) handles it very poorly. Plus, the book is in an old format, with two columns, so for the text that does scan correctly, I have to go by hand to separate the lines of text.