My understanding is that Word Wise is based on offsets of words within the raw html of the book. For the words to match up you would need to duplicate all of the internal markup of version for sale by Amazon.
It would be more practical for you to generate a new Word Wise database that matches the file you are sideloading. See:
Scripts for generating word wise information