Also need to decide on convention for if to include all reconstructed EV entries - all reconstructed events (including retriggers) --> will cause problems when scaling
Each spectra should have number of events it represents as a member, then included in scaling.
James raised the point that Andy probably assumes 100% efficiency
Maybe even need to create two versions of spectra class