That would significantly undercount multiple iPads behind a lesser number of NAT routers or wifi access points. On the other hand, this would also overcount people who have been roaming between multiple access points (home, coffee shop, work, etc.)
Actually it wouldn't over count these people. The way they are making the estimate is by looking at at a representative chunk of the Internet, at a specific time, and then they are extrapolating the total number of ipads from that number. I think these guys know a bit more about statistics than the folks in this blog.