It does not appear possible to use importance sampling for more than one particle type. Another user has given a detailed description in hypernews: http://hypernews.slac.stanford.edu/HyperNews/geant4/get/biasing_scoring/197.html I find the same issue with versions 10.2 and 10.3.
Thanks for point this out. I need to understand if it's a genuine bug or a design limitation. I'll investigate further.
This is almost certainly an enhanced feature which is in the plan of work for this year within the Biasing Working group. We will also endeavour to see if it can be better resolved using the Generic Biasing approach.
This is now moved to a feature request.