Driving a scanning tunneling microscope (STM) tip into a metallic surface and pulling it out afterwards results in the formation of a nanometer-sized wire (nanowire) between the electrodes. The electrical conduction measured during this process shows signs of quantized conductance in units of ${\mathrm{G}}_{0}$=${2\mathrm{e}}^{2}$/h. Due to the inherent irreproducibility of the measured conductance curves, the standard technique has been to construct histograms with a few hundred selected curves. These histograms, for gold nanowires at room temperature, have shown three to four peaks at integer values of ${2\mathrm{e}}^{2}$/h, while in a low-temperature mechanically controlled break junction study, with statistics using only 65 curves, only the first peak has been reported. A proposed explanation for this apparent experimental discrepancy has been a higher nanowire temperature arising from the higher retraction speed used in scanning tunneling microscopy measurements. However, our simple estimation using macroscopic heat transport theory produces a very low temperature rise, less than 1 \ensuremath{\mu}K. In this work, an improved statistical study is presented, where histograms built with thousands of consecutive gold contact breakage experiments at 4 K show up to seven conductance peaks. Thus, no significant differences with our previous room-temperature (RT) studies are observed, pointing to a conductance quantization behavior that is the same at low, intermediate, and high (RT) temperatures.