Thanks for the clarification on Acrobat joining lines to paragraphs - but Acrobat would have to do guesswork. I guess that Abbyy FineReader or OmniPage may be more convenient tools, but I have only little experience with them. page numbers which are useless in reflowable text). gImageReader has function of rectangular selection of specific area of Image which is OCR-ed (it may be applied to multiple pages at once) which is very useful for ridding off headings and footers (e.g. As a matter of fact any text formatting is lost (it saves plain text) but usually it takes less time to recreate text formatting from scratch than correcting formatting produced by Acrobat.

Personally I prefer tesseract (current 4.00 beta version) plus gImageReader frontend. On the other hand I agree that there are better OCR programs than Acrobat.

So usually no additional converting of line brakes into spaces is necessary (apart of typical OCR errors). Newer versions of Acrobat Pro join lines into paragraphs and in general try to save layout of original documents (as I remember Acrobat 11 Pro does). It is better to use another OCR which does not keep the line breaks as Acrobat does, if the ultimate goal is an EPUB. And don't forget to convert most line breaks into spaces, so that you have free flowing text for the Epub.

Calibre ebook reader not working