A lot of proprietary ones don't either.
They only place restrictions on the user of the code. Accepting those restrictions is the price you pay for using the code.
Many proprietary licenses place serious restrictions in addition to having to pay for the software. A lot of vendors even said you couldn't release benchmarks of their products without prior approval from them. Microsoft and Network Associates have invoked this clause to prevent embarassing results from being published. NA's terms were thrown out in court, and Microsoft has gotten a bit better now for Vista at least: you can release results for any .NET benchmarks but only if you follow their rules, which include the release of your complete source code and scripts used to do the benchmark.
I think that's fine, and in fact I wish -all- benchmark programs had to follow that rule. I have written a lot of test/exercise code, and I know how easy it is to have unintentional bias leak in.
Also, releasing source of the benchmark means that -all- the tested vendors can tweak their code for better results, instead of only the vendor who pays off the benchmark writers. [slight /sarc] The source release requirement can only improve the quality of the benchmark.
I trust proprietary benchmark programs even less than I trust proprietary applications. I use proprietary software when I have to, and I rely on it to a degree every day, but never as much as if it were open.