in reply to
Spreadsheet::ParseExcel with embedded PDF cells
As far as I understood OLE embedding and OLE Document Streams (and that never was really far), an embedded object is mostly a screenshot of the representation and a serialized opaque blob of the embedded object. The target machine hopefully has the same OLE-enabled application (that is, Acrobat Reader in your case), and that application is basically called upon to refresh the "preview" bitmap or edit the embedded object if necessary.
At least from this rudimentary understanding, I doubt that there will be a generic way to get at the "embedded file", as the data blob is mostly opaque. Possibly you can write a special case hack for getting at the PDF, as the PDF likely has a distinctive start signature (%%) and contains mostly text and gzipped blobs.