I think picking apart small details of the calorimetry is silly. Large error sources would be another story but it is puzzling what such a source could be, absent deliberate deception, especially considering the calibrations. I have not read any theories for such errors here. Did I miss it?
Not exactly a theory, but a clear evidence of a possible source of large errors in Mizuno's experiments (1): the spreadsheets of the active and control tests carried out in May 2016 (120 W in, COP=2 ca.) show the existence of serious inconsistencies about the electrical parameters of the input power. Similar inconsistencies could easily explain the extraordinary results (50 W in, 250 W out) claimed for the R19 tests.
IMO, before taking into account any small detail of the calorimetry or attempting a replication, it would be necessary to understand what caused the inconsistencies contained in the May 2016 spreadsheets and to carefully look at the data recorded during the R19 active and control tests.