Xb1 sales have been off for almost every week this year by huge percentages, not just a 'couple of months'. And the may numbers were a farce in every respect. Whatever they're sampling, they'd be better off ignoring and just guessing instead.
That said I do enjoy the site, I'm not mad at anyone, and I keep pulling for vgc. I think a reasonable target would be to achieve around a 10% margin of error vs NPD numbers. That should be doable, but would probably require an honest look at why the current sources of data are so broken. Adding more retailer sources would also help.