Just read the whole thing and I'm not at all impressed. It turns out socially engineered malware is just regular malware. The report is very low on content. The methodology isn't described at anywhere near the level of detail required to actually reproduce this. The report just throws around a few percentages and repeatedly refers to increases in protection from 8% to 17% as 9% improvement, one of my pet peeves. I guess I shouldn't be surprised, but for some reason I expected more.
The report describes an "in-the-cloud reputation-based system" for identifying URLs leading to malicious downloads (called "socially-engineered malware" in the report).
"From an initial list of 12,000 new suspicious sites, 1,756 potentially-malicious URLs were pre-screened for inclusion in the test ... Of the initial 1,756 URLs, ultimately 562 URLs passed our post-validation process and are included in the final results"
Since they don't describe their pre-screening or "post-validation" process anywhere in the report, you can't assume it's a completely random set of malicious URLs. In fact, the URLs they tested might have been cherry-picked to give IE8 a higher score.
In 2009 and 2008, NSS Labs did very similar reports with very similar results. They were paid for by Microsoft. I don't know whether this one was also paid for by Microsoft (it doesn't seem to say anywhere in its text, though I didn't look very hard).
Obviously there's reason to be a bit more skeptical about a paid-for study than a truly independent one.
Full report (PDF): http://nsslabs.com/test-reports/NSSLabs_Q12010_GTRBrowserSEM...