Part 1 fixed the export CSV, but the DAO creation code relied on hard-coded column names which changed in part 1.