Increases in studies on the network dynamics of crime groups and co-offending has led many scholars to reflect on potential measurement biases arising from a reliance on official data sources. A problem of official data is that it forces boundaries on criminal groups that are much more fluid and dynamic than they seem. Drawing from interviews with an individual embedded in a terrorist organization and court documents records, we apply longitudinal network methods to examine how the extent to which official data influences assessments of a criminal groups. Findings show that only a minority of participants interacting with the group were charged for a crime. Yet the majority had an impact on the evolution of the group. Ignoring non-criminal affiliates masks the full scope of covert groups and the variation that can assist in understanding how groups emerge and evolve.