They can make the statement by taking a random sample of DNA sequenced so far and comparing it to the human genome. Obviously that number may move as more is sequenced, but as mutations tend to be pretty randomly distributed, it seems likely that that number will hold.
Makes sense. Thanks...