Empirically, testing the error register this way also just doesn't help. At least with Hitachi/IBM disks, I get 01h ("drive 0 passed, drive 1 passed or not present") from drive 0, and a mirror of that from phantom-drive 1, so it looks like both drives are present.