Each instance has one value of pT. Remember, an instance means, for the muon variables, one muon. So in the example above event #2 has 3 muons:0,1,2, each one of them with its own pT value. The values for those 3 muons form the array.
Note that not all the variables are arrays because some variables, like the event number, number of muons, missing energy (met), triggers, etc, are just one for each event. You can’t have many missing energies in one event, for example.
The runno is the run number, but since this file uses MC simulations, the run number is not really valid; it is there for completeness.
When you draw a histogram like for “mu_pt” you are actually drawing the pT
of each single muon in each single event, i.e., you are plotting all muons pTs. However, if you plot “mu_pt[0]“, for example, you would be plotting just the pT for the hardest muon in each event.

As far as the inv. mass, you will get a single value but for *EACH* event. In other words for each event you will be calculating the invariant mass of the “two hardest muons” system. So you will get as many inv. mass values as events are in your root-uple file.
-e