In the Glowstick experiments, I found several additional sources of thermocouple error. The electrical conductivity of ceramics increasing with temperature also applies to typical refractory cements. I found that silica+alumina types were a particular problem when approaching 1000°C, when the silica component liquifies and migrates into any available void or crack. This loosens the bond, and forms a potential conduction path through the body of the cement.
Another problem I noted was the sensitivity of the thermocouple to minute changes in the thermal conduction path at its attachment point. Most cements will shrink as they cure and this can break all or part of the bond at the substrate. The thermal path to the thermocouple is thus changed from conduction to a mix of conduction and radiation across the gap, however small it might be. There may also be a subsequent variation in surface contact due to differential thermal expansion of the cement, substrate and thermocouple metals. It may be even more pronounced in this experiment, where the thermocouple is attached directly to the metal fuel tube.
This has proved to be a difficult engineering problem, resulting in rather large error bars for the temperature measurement, as much as ±50°C at 1000°C operating temperature. As a consequence, measurement of thermal gain less than 1.1 is not possible. I would want to see greater than 1.2 in my experiments before claiming anything. I'm working on better thermal coupling technique and welcome any suggestions.
AlanG