The experiments and simulations are the ones weak on statistics, by not differentiating between events requiring random chance and events that allow for weighted probabilities.
Dembski and Behe maintain the flagellum, being irreducible, cannot be formed using a mechanism of weighted probabilities, but only truly random probabilities.
> Dembski and Behe maintain the flagellum, being irreducible
Which it's not.
http://www.millerandlevine.com/km/evol/design2/article.html
Since they are wrong from first principles, the rest of their arguement can be dispensed with.