Let's be clear, the article is comparison between frameworks. The benchmarks should therefore reflect the comparative speed of the same web application written using both rails and snap. I'm not aware of any attempt to do this, so I attempted to approximate it using the speed of the language in general and you can see that from the results I posted that the slowest Haskell GHC is faster than the 25% percentile of MRI ruby 1.9. This is a useful observation; A simple pong benchmark has absolutely no bearing whatsoever.
I found some on the Snap framework web site[1], but they may be a little out of date.
[1]: http://snapframework.com/blog/2010/11/17/snap-0.3-benchmarks