In many experimental contexts, whether and how network interactions impact outcomes of both treated and untreated individuals are key concerns. Networks data is often assumed to perfectly represent the set of individuals who might be affected by these interactions.