Uncertainties due to the magnification variation in sub-nanometer metrology using transmission electron microscopy (TEM) were experimentally evaluated by comparing the measured values of the (220) lattice spacing of a crystalline Si specimen acquired under various conditions. Interday variation of the magnification, intraday repeatability, rotation of the TEM image, specimen exchange, specimen position, defocusing, magnetic hysteresis of the lenses, projection lens distortion, and measurement errors were considered as the uncertainty components. The obtained results reveal that the major uncertainty components are the interday magnification variation, intraday repeatability involving nonuniformity of the specimen structure, magnetic hysteresis of the intermediate lenses, and projection lens distortion. Among these components, it is expected to be feasible to suppress the interday magnification variation to a negligible level by daily magnification calibration, suggesting that minimizing the uncertainties due to magnetic hysteresis of the intermediate lenses, projection lens distortion, and nonuniformity of the specimen structure will be the key factor for further reducing the uncertainty of sub-nanometer metrology using TEM. Furthermore, we found that magnification calibration using an appropriate reference material (e.g. the lattice fringes of crystalline Si) before every measurement should enables sub-nanometer metrology with a relative uncertainty of 3.2%, even if the specimen is introduced into the TEM system by specimen exchange after magnification calibration.