That's a good point, though I think it's unlikely that 5% of the sites and 95% of the sites both make up ~50% of impressions each.
But to your point, we don't know unless we know, and my assumption could be totally wrong.
But to your point, we don't know unless we know, and my assumption could be totally wrong.
1) Pareto: where there's a power law based distribution. So it is possible that 5% of sites have more than 50% of impact.
This just means it isn't out of the bounds of possibilities. But we'd have to dig deeper into which sites garner the majority of eyes (though we'd expect a power law in this).
2) Left learning areas tend to be more population dense.
Of course with these two considerations we haven't really made any conclusions. I just want to point out that a small number of sites can reasonably hold the majority of viewership. Which sites those are and which bias they entail, that requires a real analysis. We've just stated that the conjecture isn't impossible.