What I'd suggest in your case is to do a simple node query to get the node and document data without the coupled columns. This makes your query generic and doesn't cause issues later on with the multiple page types. Once you get your 2 lists of node or document ID's, you can do your intersect and get a list of ID's without all the extra data. After you have your list of ID's do a query to get your full data with coupled columns.
There could be an easier way, but this should work until you find that solution.