The message I got wasn't related to the number of addresses affected (though I've been using this approach for a couple decades), but IIRC regarded whether the datasets in question were free.