Why isn't Oops Detection done within the manuscript like editing and proofreading?

There are several reasons, but the most important ones are these:

• We do Oops Detection on a Kindle. This lets us check for formatting issues if the project is sent over in a mobi format. If it's not, I can convert it, though we can't check for formatting issues in that case.
• Editing a document in Word often introduces errors as a byproduct of Track Changes. Spaces get added or deleted, a change gets accepted instead of rejected, etc. To prevent this, it's best to do those last fixes manually within your working document.