I also tried playing with capacitors on that circuit a while back with little success. I still think the best way to solve it should be with better software... something that would be more flexible when looking at the signal coming in. Some type of algorithm that knows what the wave should look like and have a good guess where the next edge should be coming in at, based off how fast RPM's could possibly change between cycles. I bet you could easily "throw out" a good chunk of spikes that way.
