Keeping zero count data

[jfkcooper]

For polarised data/simulations it is actually much better if a simulation will always have the same (and expected) number of points. We currently remove zero count points, but probably shouldn't, for this and other reasons