As targeted, single-conception curriculum research becomes more prevalent in physics education research (PER), the need for a more sophisticated statistical understanding of the conceptual surveys used becomes apparent. Previously, the factor structure of the Force Concept Inventory (FCI) was examined using exploratory factor analysis (EFA) and exploratory multitrait item response theory (EMIRT). The resulting exploratory driven factor model and two expert-proposed models were tested using confirmatory factor analysis (CFA) and were found to be adequate representations of the FCI's factor structure for postinstruction student responses. This study complements the literature by using confirmatory multitrait item response theory (CMIRT) on matched preinstruction to postinstruction student responses (N = 19745). It was found that the proposed models did not fit the data within a CMIRT framework. To determine why these models did not fit the data, an EMIRT was performed to find suggested modifications to the proposed models. This resulted in a bi-factor structure for the FCI. After calculating the student ability scores for each trait in the modified factor models, only two traits in each of the models were found to exhibit student gains independent of the general, "Newtonian-ness" trait. These traits measure students' understanding of Newton's third law and their ability to identify forces acting on objects.