Hi,
I believe if you have a "template" table selected as the source in the xls reader, it will use this as a guide for the column structure for the files read in using the flow variable. As long as the template contains all possible columns (and the headers and column types match) it should work. I have had to do this, but I think it was with the csv reader?
Alternatively, you may be able to have your loop read the template file first (you may need a row with dummy data), delete the data later.
Hope this helps.