Problem is that friction (and the physics around it) are not something that can be broken down into a percentage easily...at least not in anything that translates directly from real world fact to Physics Engine practice.

Fudging the numbers is what you need to do to get the results that you want...forget what the value is on paper or what the numbers say...but ask your self "Does it look right?"

This is important because stuff like rubber on concrete has a very high ammount of friction. Especially when the rubber is a bit sticky...but coefficients only hold true in a perfectly controlled lab environment. However when that rubber gets heated up too much, it becomes very slick (real world example...locking up the tires till they are sliding...melts the rubber and you are hydroplaning on a puddle of molten rubber...leaving balck marks all the way). Ice on ice will actually lock up in most cases... However for a moment it will slide quite well.

Whats the point? There is just too much going on that can not be taken into account in simpified physics simulations to use real numbers. You have to make stuff up as you go in order to achieve realistic results without needing a super computer to process the calculations.