This doesn't explain, how the COP could change during experiment and even how it could rise, when the temperature of reactor also raised.
The increase of temperature would lead into increase of thermal loses and if they got systematically underestimated (because of lower emisivity or nonplanar geometry of the reactor), then their increased underestimation should also lead into lowering of COP - which didn't actually happen.
Zeph -
You have been consistently and strongly saying there is much evidence Rossi's stuff works, those who don't see this are willfully blind, etc.
I am happy to consider this evidence, with you at length. So that any blindness can be considered. It is important that we be careful doing this because as I'm sure you know anecdotes and web facts (not to mention now alternative facts) often look different when examined in detail.
I'll start with your impressive graphs above. First, I'll agree with you. They are impressive. When you first look at the Lugano evidence, although you might reckon there are a bunch of mistakes that could account for the claims, this acceleration in COP where 100W in turns into 600W out, is quite something.
As Paradigmnoia says, the numbers show this to be an artifact but i expect you think he is willfully blind. Certainly P has not now justified his position, though as he and I point out there are hundreds of pages here doing exactly that. In many of which P was arguing your side. So this is a matter that has been hotly debated and the consensus now reached amongst those who stuck through that debate might count for something.
I'm going now to give as simple an answer as i can to this X6.
Firstly, the known issues about emissivity account for the approx X3 COP, which turns 800W into 2400W. The output power is indirectly measured via temperature and scales as T^4. The temperature was massively over-estimated via its indirect measurement and the problematic Al2O3 surface used.
Secondly, and this is the issue I will address at length here, the way that thermography error operates is quite subtle and leads to surprising results. In fact it leads (using the figures from the Lugano report, but correcting the errror) exactly the COP acceleration that is shown in your graphs.
If you want to validate this it is not so difficult:
Take the code from TC's report (as found on lenr-canr) and run it, to generate numerical results. There are various fudge factors which can move the COP up or down by 20% or so. The code as given shows this by printing COP figures bracketed with different values of fudge. BUT - these don't affect the COP acceleration (because most have nearly the same affect at the two test temperatures). What is remarkable is that this acceleration goes away when COP is correctly calculated and the two COPs become identical to within 0.5%. This lack of change is resilient as you alter the various fudge factors. BTW what I mean by fudge factors is that you need a few assumptions to get any results from Lugano. Change those assumptions you alter the COP results. But not the COP acceleration. The Lugano authors were aware of this which is why the noted this acceleration as particularly significant. Except it is not because it is an artifact when things are calculated correctly.
So that is the numerical answer, based on Lugano figures. Can it be explained theoretically. The key to understanding this is that total emissivity in Al2O3, changes rapidly over the range of temperatures calculated in Lugano here. The report uses this change in total emissivity (because they feed it into the Optris device) to calculate temperature. That gives an error because ove rthe same range band emissivity does not change and stas close to 1. Therefore the inflation of temperature their wrong method generates depends strongly (and nonlinearly) on the temperature - because it comes from the difference between total and band emissivity which is much larger at higher temperatures than low.
it is in reality not COP acceleration - but "Levi thermography artifact acceleration". (I attribute this to Levi since when last heard of he was still to Mats robustly defending this flawed method - he seems very invested in it).
PS - I realise you quote other evidence. And that this other evidence must also be examined in detail. Could we however first please deal with Lugano? I'll happily lead you through either running the python code, or any of the theoretical equations it embodies.