The source inversion model used in the Biological agent Event Reconstruction Tool (BERT) computes the potential release location(s) based on air sampler dosage measurements obtained from filters, wind measurements, and estimates of atmospheric stability and boundary-layer depth. BERT utilizes a fast running diagnostic wind model, a segmented Gaussian plume model, and a source inversion methodology called the collector footprint approach. All are simplifications compared to reality and so evaluation of the source inversion scheme is necessary to understand where the approach works well and where it doesn't. In this evaluation study, we have utilized SF6 concentration measurements obtained during the multi-agency Joint Urban 2003 outdoor field experiment held in Oklahoma City. Point source releases were conducted in the central business district and a large number of measurements were obtained up to 4 km downwind. The concentration measurements from the continuous releases are used to back calculate where the source region is located and to provide bounds on the release amount. A number of indicators have been identified to assess BERT performances. In this paper, we present an overview of the model performance broken out by number of collectors hit, day vs. night releases, and time variable vs. steady wind cases.