Actually, first i was doing the first.
Because it was not good enough on some occasions, Then I added code to do the second.
It seems in some situtations the first brings the best results, in other situations the second.
The sort of feeling I have if the morph is local and small then applying same delta to both vertices is best. If its a full blown deformation of the whole body then the first approach is better. But maybe my example of a FBM is too extreme (i morph a pose into another pose). If I look at the morphed figure in the zero pose that is really looking odd.

I had hoped there is kind of a theory that clearly determines what has to be done.