I was thinking more of 42 columns as in your original spec. 700 columns just adds to the complexity
However I am intrigued why you have the documents combined in the first place. What process do you have in place to create such a combined document?
On the face of it you appear to have a number of question sets combined into a single document. It would surely make things simpler if you were to have one question set per document, which makes it easier to extract the data by document to a separate line of the workbook (which is what I thought you were doing at the outset, when I suggested the add-in).
It might be possible to split the document into those separate documents, and then extract the data, but it would help to know both why you have it that way in the first place, and how the datasets are separated in the document so that they can be logically split into separate documents each with the same field sets.