The performance of over 5000 students in introductory calculus-based mechanics courses at the Georgia Institute of Technology was assessed using the Force Concept Inventory (FCI). Results from two different curricula were compared: a traditional mechanics curriculum and the Matter & Interactions (M&I) curriculum. Post-instruction FCI averages were significantly higher for the traditional curriculum than for the M&I curriculum; the differences between curricula persist after accounting for factors such as pre-instruction FCI scores, grade point averages, and SAT scores. FCI performance on categories of items organized by concepts was also compared; traditional averages were significantly higher in each concept. We examined differences in student preparation between the curricula and found that the relative fraction of homework and lecture topics devoted to FCI force and motion concepts correlated with the observed performance differences. Limitations of concept inventories as instruments for evaluating curricular reforms are discussed.