The simulation would be fixable by including an “in a group” factor and getting some information on actual boarding behaviors into the simulation. That would require (gasp!) experiments to measure those things. How about if an airline tried several different methods of boarding to see which worked, I’m sure it has happened already.
It would be a simple thing to enter different zones on the boarding passes. The gate agent would always do the same thing, the cards would have different orders on them and someone needs to time each one. Do thousands of these over a bunch of different times and conditions and you are set.
