I recommend using the Text mode for text recognition. It does a decent job there - better than One Note, especially if we consider the recognition of printed & scanned documents or documents of a language other than English.
The accuracy drops with screenshots or other low resolution camera images, but it is quite good with most scans. In my experience, it works well however with PDF files that contain no accessible text (which may come handy if the edit mode is locked...).
The formatted mode only works well if the document is a conventional flow text with headers, paragraphs, etc. It does handle the case of multiple columns well, but has issues with stuff like background pictures (like magazine frontpages) and tables. It may or may not get it right, depending on the document layout. I am not sure about the performance issues mentioned above, it works quite fast for me right now.