High-resolution tracking of gold nanoparticle-labeled proteins has emerged as a powerful technique for measuring the structural kinetics of processive enzymes and other biomacromolecules. These techniques use point spread function (PSF) fitting methods borrowed from single-molecule fluorescence imaging to determine molecular positions below the diffraction limit. However, compared to fluorescence, gold nanoparticle tracking experiments are performed at significantly higher frame rates and utilize much larger probes. In the current work, we use Brownian dynamics simulations of nanoparticle-labeled proteins to investigate the regimes in which the fundamental assumptions of PSF fitting hold and where they begin to break down. We find that because gold nanoparticles undergo tethered diffusion around their anchor point, PSF fitting cannot be extended to arbitrarily fast frame rates. Instead, camera exposure times that allow the nanoparticle to fully populate its stationary positional distribution achieve a spatial averaging that increases fitting precision. We furthermore find that changes in the rotational freedom of the tagged protein can lead to artifactual translations in the fitted particle position. Finally, we apply these lessons to dissect a standing controversy in the kinesin field over the structure of a dimer in the ATP waiting state. Combining new experiments with simulations, we determine that the rear kinesin head in the ATP waiting state is unbound but not displaced from its previous microtubule binding site and that apparent differences in separately published reports were simply due to differences in the gold nanoparticle attachment position. Our results highlight the importance of gold conjugation decisions and imaging parameters to high-resolution tracking results and will serve as a useful guide for the design of future gold nanoparticle tracking experiments.