Whether that effect is missed or not by the calculations, it doesn't come into play when comparing sporting events to other sporting events.
Not completely correct. I will not watch a minute of any golf tournament as I find it uninteresting. Likewise, very few bars are attracting large audiences for those events. Nevertheless, said events will garner large shares and volume of media coverage that may surpass events such as Final Four of even World Cup (in the US). To say a sport that attracts a very narrow demographic is more popular (when the definition of popularity is broad appeal) is undoubtedly incorrect. I am of course ignoring the purpose of measuring these data - namely the golf audience is substantially more valuable than the World Cup audience in the US, the former well modeled by advertisement community and the latter too diverse to model. (Look at the ads for golf - investments, insurance, luxury and Veblen goods; they know who they are chasing or trying to influence.)
In science we measure things with the recognition that accuracy and precision and not synonymous. Actions one takes to increase precision (number of tv's tuned to an event) may not not increase accuracy (number of viewers for said event). They could weight the location of tv's (ones in a commercial district more likely to have multiple viewers) or include alternate data (Twitter, page hits to related web sites) to account for this divergence, and probably do. But the purpose is to influence advertising purchases not to measure involvement by the populace, so there's not that much interest in accuracy or precision, as much as there is creating confidence in the data they provide.
