By "official" I think he meant the benchmarks done by the websites themselves, using proper methodologies. In this article the benchmark was submitted by a user, using synthetic tests from a single application.