I'm perplexed as well. That the perfect mixing models most people think of for infectious diseases are an imperfect approximation of the actual dynamics on a graph is well known in epidemiology.