Hi Christian,
What if you store all the distributions in separate files and let the energy and angles being sampled via those files in a macro? If yes, see my response here.. This is implemented via GPS, which I highly suggest as it is straightforward to implement.
With ParticleGun I have always generated random numbers from 0 to 1 and defined the angular range using those numbers. For example, for each event a random number is sampled, hence a random angle.