Super Bowl XL households second only to M*A*S*H finale? WTF?

This is why I've always taken a network's ratings claims with a huge grain of salt: According to this AP article, Super Bowl XL was seen in 45.85 million homes, second only to the 1983 finale of M*A*S*H in the history of TV. However, the rating itself, 41.6, was only up slightly from last year's game (probably because a rating point represents more homes than it did last year or in 1983). Also the share, which measures the percentage of TVs in use, remained the same at 62.

Got all that? But wait, there's more...

141.4 million people watched the game at any single point, which was second only to last year's game in the history of TV. So, even though the rating was higher than last year's game, the number of people watching at any particular point was lower. However, the average number of people watching throughout was higher than last year's game (90.7 million people to 86.1 million), the highest number since Super Bowl XXX in 1996.

So, class, in conclusion: more homes watched the game than any other program in history except one, but less people watched at any one point than even last year's game. Finally, more people watched the game on average than last year's game, but still less than the number from 10 years ago.

Wow, that's some fine statistical hoop-jumping on the part of the folks at Nielsen. I mean, I have an engineering degree and I can't keep track of all this. My guess is the number of people watching in bars and at parties skewed these numbers a bit, but really I can't explain it more than that. If you have any idea how the numbers came out like this, let me know in the comments.