That's a very astute observation, one that I've personally seen overlooked several times by people using that data point to support their argument. That kind of statistical begging-the-question annoys me to no end.

The number I've heard from people who quote it is all over the map, sometimes it's 10%, or 20%, even as high as 50%, but I've not once seen a citation of any of them. So I'm inclined to treat that statistic as BS unless proven otherwise. But as you say, even a legit number is probably misleading due to the circumstances surrounding the data gathering.