There’s the fact that the experienced internet users are unlikely to install the Comscore kind of software.

Shared computers are a big problem, too: Comscore keeps registration of the characteristics of the user who registered and installed the software. That means the usage patterns of families with shared computers cannot be reported on reliably. However, the fact is that there isn’t really a good way of measuring this kind of thing. Hitwise gets fairly accurate traffic data by buying traffic logs from ISPs, but that can’t show the characteristics of the individual users. However, I tend to trust the Hitwise traffic data more than other sources for general traffic trends. I tend to regard Comscore & Alexa data as “possibly indicative but not definitive, and always requiring further analysis”.

Comscore is a media measurement site similar to Hitwise or the Alexa site rankings. They do a decent job, but people are beginning to take their data way too seriously for the accuracy that their methodology delivers.

Essentially, they (like Alexa) rely on users installing software which monitors which sites they visit.

That methodology means your samples are not (and cannot be) representative of internet usage.

For instance, most workplaces, schools and colleges have policies and security settings prohibiting the installation of this kind of software. That’s already knocked out a very large proportion of internet users from the sample.