Please look at Figure 20 of T. Mizuno / Journal of Condensed Matter Nuclear Science 25 (2017) 1–25 (ref 1 of the ICCF22 preprint). It shows the inlet and outlet T’s on a run from the same calorimeter. Two points from that Figure: (1.) The inlet temperature tracks the outlet temperature in the first (major) part of the trace, (up to a little past 7 hours) and (2) the noise levels of the two signals are slightly different but remain the same throughout the whole run. Now look at the Figure I posted. The purple curve clearly shows regions of increased noise levels that don’t track the output temperature, and the noise level of those regions is more like 0.8-0.9 degrees, nor <0.1. In fact the noise levels near the end of that run show ~.2-.3 degrees span, so they change during the run. I can see the inlet T noise in Fig. 20 might be the 0.1 degrees, but the outlet is larger. What is most important though is the loss of tracking that we see in the spreadsheet plots. That indicates an ‘external’ source of signal that causes a dip in Tin, which translates to a positive jump in Wout. One common cause of that behavior is electrical noise.
Room temperature drift should really not be a problem. That’s why one subtracts the Tin from the Tout instead of assuming a constant Tin, to correct for drifts. But as Ruer noted, one likes the Tin to be pretty unchanging, certainly not tied to the outlet T. That is a problem. (BTW, the Ruer quote is from page 58 of ICCF22_Abstracts.pdf.)
Does this invalidate the calorimetry? I don’t know. It is a red flag is all.
Replication will answer all.