The problem isn't the version of Word, it is in the nature of OCR. OCR is a complex process and pays no attention to how Word works. When I want to actually use a document in Word that has been scanned, I will usually save it as plain text (.txt) and then copy that text into Word. I then use styles to format the text.

This is a fair amount of work, but nothing compared to dealing with anomalous formatting created by the OCR process. (Some OCR programs are better than others, but none produces a document that edits like one that has been directly produced in Word.)