Advances in cryo-electron microscopy instrumentation and sample preparation have significantly improved the ability to collect quality data for biomolecular structures. However, achieving resolutions consistent with data quality remains challenging in structures with symmetry mismatches. As a case study, the bacterial flagellar motor is a large complex essential for bacterial chemotaxis and virulence. This motor contains a smaller membrane-supramembrane ring (MS-ring) and a larger cytoplasmic ring (C-ring). These two features have a 33:34 symmetry mismatch when expressed in E. coli. Because close symmetry mismatches are the most difficult to deconvolute, this makes the flagellar motor an excellent model system to evaluate refinement strategies for symmetry mismatch. We compared the performance of masked refinement, local refinement, and particle subtracted refinement on the same data. We found that particle subtraction prior to refinement was critical for approaching the smaller MS-ring. Additional processing resulted in final resolutions of 3.1 Å for the MS-ring and 3.0 Å for the C-ring, which improves the resolution of the MS-ring by 0.3 Å and the resolution of the C-ring by 1.0 Å as compared to past work. Although particle subtraction is fairly well-established, it is rarely applied to problems of symmetry mismatch, making this case study a valuable demonstration of its utility in this context.