Hi Steven,
I prefer always to use contact as I so not like the separation springs as they can mislead; particularly if a hole is partially separated.
I don't use account for stiffness as it is slower when one has to do several analyses starting from the same preload; it is better to perform a preload study and modify the preloads by hand. (this has always been my criticism of this functionality, fastener preloads are not automatically updated)
I only began using the fasteners with separation springs in this case for speed (time constraints). But I got stupid answers. Because I cannot trust the answers and have reverted to contacts where the answers are much slower but realistic. This is a connection where I know there will be no separation (preload is sufficient).
The model I showed in my first post (and have attached here) was just to test. The sign of the separation stress should change if the components separate and I now believe there to be a bug.
I haven't had time to try other versions/cut and the model attached is one I will submit to ts.
Creo3.0 M030
Regards