This was a tough number to estimate. I can see how Nielsen is not random in terms of returns, and it may bias their number in favor of cable viewers. They send you a booklet for every tv you have, but it is tempting not to include tvs you know you won't be watching that week. So I'm not sure Nielsen gets an accurate picture of how many OTA homes their are, and how many tvs per OTA home.
Another place underestimating can come in is in their allowing coupons to be given away. Even if they were right on the 2.1 tvs per home, the estimate is off by 50%, if everyone has a friend order them a coupon so they can have a box for their VCR, too.