Currently, when relaxing helix rolls, all inter-helix crossovers are accounted for, but actually the calculation of the relative angle from one helix to another (based on the position of their 0 offset) only makes sense if the helices are in the same helix group, i.e., are parallel. If they are in different helix groups, we should do one of two things (possibly configurable with an option):

1. (easy) ignore those crossovers
2. (hard) do a more detailed calculation of the angle of the other helix, based not only on the helix's origin position, but also on the offsets of the two points on the helices the crossover connects.