>It wasn't 24%, it was 24% of the half that could be identified.

with a large enough data set the chance that the half that cannot be identified is also 24% looks pretty good, unless there is something specifically about these brands that makes the identifying of them easier than other brands - which is also a pretty good chance I would say.

If there is a good likelihood of identifying these brands specifically then I guess it is actually pretty close to 12% (assuming there are still some small amount which cannot be identified)