How to build Hierarchical Clusters?

Hello,

I know the possibility of using linkage(X,'single','cityblock') and cluster(). My problem is that the btree cluster output is nondeterministic when different input data is used. That is quite logical of course. However, in my special case it would be very convenient to know that on the lowest cluster level the distances between the points is exactly zero, on the next level 1 and so on, no matter how homogen or heterogen the data is structured. This may lead holes but this is no problem here.

Is there a possibility to tweak the cluster() function to work this way?