I discount Goat's hypothesis for the following reasons: As shown in figure 10 they split the reactor IR camera image into 10 segments plus the ends. They record the temperature for each segment. As shown in the photograph, some segments were incandescent and others were not. If incandescent segments showed up erroneously being much hotter than the other segments, I suppose they would notice this discrepancy. Alumina has "good thermal conductivity:"
http://accuratus.com/alumox.html My point is, if the surface really is ~750°C (meaning there was no excess heat), that temperature would show up in the dark segments. The incandescent segments would show up as considerably more than 1200°C, to make the average around 1200°C. Such a huge temperature difference is not possible. They would know that is a bogus reading.