Bifacial photovoltaic (PV) modules have greater energy yield than traditional monofacial modules because they convert front and rear incident irradiance to electrical energy. Single-axis tracking systems can further increase energy yield and reduce the levelized cost of energy by rotating the modules throughout the day. However, racking elements in tracking systems introduce both shade and reflections on the rear face, increasing irradiance nonuniformity and fostering further electrical mismatch that reduces module power. The impact of racking, particularly that of the torque tube which spans the middle of the rear collector surface, must be accurately quantified in energy yield predictions to increase stakeholder confidence, and hasten the adoption of tracked bifacial modules. Isolating the torque-tube-reflected irradiance incident on the modules is crucial for this work. This is achieved by implementing arbitrary two-dimensional (2D) irradiance sampling on a module under test in bifacial_radiance, a ray tracing bifacial PV model, and taking the difference in rear irradiance profiles for simulations with a reflective and an absorptive torque tube at each timestamp. We calculate the TT reflection for the central one-in-portrait (1P) and two-in-portrait (2P) modules on horizontal single-axis trackers over hourly timestamps in a typical meteorological year in Livermore, California, USA. We introduce the TT reflection 2D irradiance profiles as additional light sources in DUET to quantify the TT reflection's impact on irradiance, electrical mismatch, and energy yield, as well as an incidence angle modifier's impact on TT reflection. We analyze the TT reflection based on sun zenith and diffuse fraction to group consistent illumination conditions across the year. We identify that TT reflection reduces electrical mismatch by partially offsetting TT shading, and increases annual energy yield by 0.11% and 0.18% in our particular 1P and 2P systems. While the overall impact of TT reflection is greater in the 2P system due to direct beam light incident on the TT, the TT reflection's greatest instantaneous relative contribution to total energy yield is larger for the 1P system, at high diffuse fractions and sun zeniths. For future work, we recommend validating simulation results with and without TT reflection against experimental data. The simulation method used for isolating the TT reflection may also be repurposed to help inform new TT designs that minimize electrical mismatch. Finally, instead of relying on bifacial_radiance to isolate the TT reflection, we recommend incorporating the TT as a Lambertian reflective surface in 3D view factor models with detailed shading for further TT reflection simulations.