There are many critiques of theory, or based on theory. I don't know much about them, but they are all wrong,
Well, actually...
I informally here provided a detailed theoretical critique of the proposition that Ed's calorimeter could safely have errors due to heat variation on the inside of the box bounded << 0.5% based on the 3 tests described in his paper. That theory related to the fact that, in principle, to balance 12 Seebeck effect panels you need 11 (or 12 if calibrating as well) orthogonal measurements.
I have no intention of writing this up formally here - it is basic maths, but I could do so. The formal proof would look at the number of independent linear equations needed to determine uniquely 12 unknown coefficients.
Notwithstanding that, practically - such calibration is not needed. But more calibration than has been done is needed if an accuracy of << 0.5% is required to be proved, and it is pretty simple. note that ed has released no information on the relative matching of the Seebeck effect transducers, nor their model, so this could be looked up. Without that it is important to be pessimistic.