1. Although NetBSD did perform better on many of the benchmarks (kudos
to them), but the difference is usually not dramatic.
2. The benchmarks are strictly uniprocessor benchmarks.
3. The author does admit at the end of the article that NetBSD still
uses a big, giant lock around the kernel and the benchmarks might be
very different on a multiprocessor system.
4. The benchmarks are only NetBSD versus FreeBSD, so it's hard to judge
where the performance of each system fits in the grand scheme of things.
It could be that both systems are performing very well.  The
benchmarks need to include at least one non-BSD system (usually Linux)
in order to get some perspective.
5. It would have been nice if FreeBSD 4.10 and NetBSD 1.6.2 were also
include, so we could see the relative progress (or lack) of each system.