Systematics: catch the case where a variation does not affect the selection itself

Closes #44 (closed)

A bit of background: systematic variations (e.g. "jerup") are grouped ("jet") for reasons of efficiency (if one of the jet variations is found in an expression, there's no need to check for all of them, we probably need to do variations, and it's checked later what exactly). The code assumed that if a group of systematics affects a selection, all its variations affect the weight. The counterexample that triggers #44 (closed) is that the unclustered energy variations of the MET do not change the jets.