High-accuracy, short-range distance measurement is required in a variety of industrial applications e.g., positioning of robots in a fully automated production process, level measurement of liquids in small containers. An FMCW radar sensor is suitable for this purpose, since many of these applications involve harsh environments. Due to the progress in the field of semiconductor technology, FMCW radar sensors operating in different millimeter-wave frequency bands are available today. An important question in this context, which has not been investigated so far is how does a millimeter-wave frequency band influence the sensor accuracy, when thousands of distance measurements are performed with a sensor. This topic has been dealt with for the first time in this paper. The method used for analyzing the FMCW radar signal combines a frequency- and phase-estimation algorithm. The frequency-estimation algorithm based on the fast Fourier transform and the chirp-z transform provides a coarse estimate of the target distance. Subsequently, the phase-estimation algorithm based on a cross-correlation function provides a fine estimate of the target distance. The novel aspects of this paper are as follows. First, the estimation theory concept of Cramér-Rao lower bound (CRLB) has been used to compare the accuracy of two millimeter-wave FMCW radars operating at 60 GHz and 122 GHz. In this comparison, the measurement parameters (e.g., bandwidth, signal-to-noise ratio) as well as the signal-processing algorithm used for both the radars are the same, thus ensuring an unbiased comparison of the FMCW radars, solely based on the choice of millimeter-wave frequency band. Second, the improvement in distance measurement accuracy obtained after each step of the combined frequency- and phase-estimation algorithm has been experimentally demonstrated for both the radars. A total of 5100 short-range distance measurements are made using the 60 GHz and 122 GHz FMCW radar. The measurement results are analyzed at various stages of the frequency- and phase-estimation algorithm and the measurement error is calculated using a nanometer-precision linear motor. At every stage, the mean error values measured with the 60 GHz and 122 GHz FMCW radars are compared. The final accuracy achieved using both radars is of the order of a few micrometers. The measured standard deviation values of the 60 GHz and 122 GHz FMCW radar have been compared against the CRLB. As predicted by the CRLB, this paper experimentally validates for the first time that the 122 GHz FMCW radar provides a higher repeatability of micrometer-accuracy distance measurements than the 60 GHz FMCW radar.