No, I do not think plasma heating or any other heater-related noise could lead to the output temperature variations. If you look at my first graph above (just edited to fix the horizontal scale), the large output temperature variations last for more than 10,000 seconds (> 2.7 hours) after the heater is turned off.
Good point. So something temperature dependent. The noise amplitude is maybe closely proportional to the reactor temperature delta from ambient? What about instability in the calorimeter airflow leading to modulation of the output temperature (depending on how much air goes past the hot reactor, and how much avoids it). That has nowhere been considered and would give these results. You can quite easily get surprisingly large instabilities in flow systems.
The calibration and active heating (and other stuff maybe) is different - e.g. internal plasma vs external heater - so nothing can be concluded from this happening in one case only.
EDIT - I note that ascoli (below) has the same view as me on this - citing turbulence - and in this matter is probably better informed!