I summarise the issues here in my paper under the "caveats" section: specifically alumina transparency. While I agree with this comment that there are unknown errors here I disagree with the implications that (a) said unknown errors are known positive and (b) said unknown errors are potentially very large.
I also think that looking for unknown errors as validation of extraordinary LENR is itself a weird business. The desperate attempt here is to find some wiggle-room to allow Rossi's device to work, and the Profs measurements be valid. Well, the Profs chose a very indirect and flaky way to measure power out so of course there is the possibility of error. In principle the results bound the COP in the range 1.5 - 0.7 (these are conservative figures).
Although at higher optical frequencies the alumina becomes transparent the equation for power out at these frequencies is not positive. Whatever surface actually emits the power will have a smaller effective surface area than the reactor, so even if the power / unit area is higher this is counteracted. What the power per unit area is we do not know: it could be higher from Inconel wire, if oxidised. It could be lower from Inconel wire if not oxidised (the comparator emissivity, at 700C, is about 0.7). Let us suppose the wire has max possible emissivity 1. Then we have a 50% excess from emissivity, and a (50%?) deficit from the area reduction. Note that the wire will not be much hotter than the alumina surface - though it will be a bit hotter. From the equations for heat from radially from an inner cylinder to an outer one, and the thermal resistance of alumina, you can get a rough approximation. I'd expect around 50C or less. I have not bothered to do this calculation precisely (I'm remembering similar ones I did for how much hotter a core with heat production would be than the surface). Somone else could. If we take 50C that is an additional (1020/970)^4 = +22%.
Furthermore, at 700C, a lot of the radiation is at frequencies where behaviour is known and the calculation stands. So this ballpark calculation tells us that there is an error, but it could be + or - and it is probably overall less than 20%.
Finally, I agree with the poster that such an effect would be highly nonlinear versus temperature. That would lead, were it significant, to different COP values measured at 800W and 900W. However my vanilla calculation shows the same COP for these two different powers and this means any highly nonlinear error must be fairly small, further validating my calculation. One reason I wanted to release the code is for other people to try to break the near precise equal COP in the two active tests. That is strong evidence that there is no nonlinear LENR effect, no nonlinear unknown error mechanism. I did not cherry-pick this coincidence and it was robust as I changed the various parameters that affect things.