What this shows is that even the simplest sounding experiment is fraught with possible sources of error, and can be done many different ways. For the first run, I would take a glass beaker, pour distiiled, room temperature water in it, place an RF thermometer in it, record the temp, put it in the freezer and time how long it took for the thermometer to signal it was 0 degrees C. For the second run, I would then let the water thaw, microwave it for one minute, put the thermometer back in, record the temp, then put it in the freezer and time how long it took to reach the initial temp of the first run, and how long it took from that point to another zero reading.
So I would be using the same water, container, thermometer, and freezer for both runs. This would eliminate any variability you might otherwise get using two separate water samples and containers, and determining when the ice started forming.
No doubt you can find a loose variable in my methodology as well, but controlling all the variables is very different, and why different results are often obtained for any given reasearch.