There's really a tendency in this newsgroup to point people to huge documents, when a small copy and paste would make the answer so much more accessible ;-)
Anyway... so reading through the paper, it looks like its using a priority queue? Which basically is changing the algorithm somewhat compared to the C# version.
Anyway, if you can provide a working Haskell version, I'll be happy to run it. I sortof suspect that if it gave results within 30% of the C# version someone would already have done so ;-)