For what it is worth, even though the device was divided into sections, the sections are close enough to each other in reported heat that the main device can be simply modeled as a two cylinders, one 4 cm in diameter and 8 cm long, and the other 20 cm long by 2.3 cm in diameter.*
The rods are a pain, due to the heat gradient and there really isn't a good cheat for that except just calling them whatever heat the report claims they are and accepting that part is possibly a source of error. But the rods are not a major source of error, overall, due to their low average temperature and therefore heat in comparison to the rest of the device. (Maybe 4 out of 20 segments might contribute 80% of the heat attributed to the rods.)
*Two days are lumped together for each file anyways, so a little more uncertainty in simplifying the model can't hurt. For that matter, it would be interesting to see what the variance of a typical two day long file is for the various measurements.