1) the Shanetsu insulation used in this paper is rather poor relative to the state of the art insulation (Vacuum insulated panels). This product (https://www.turvac.eu/0/Products/WhatisVIP.aspx) gives extremely low thermal conductivity (3~10x the product Mizuno used) (3,5mW/mK), with just 20 mm thickness, U value less than 0,22 W/(m2K) can be reached. This would allow more precise calibrations at higher temperatures and more capture of the reactor heat in the air flow.
2) for replicators with sufficient funding a mass flow meter such as this (https://www.sierrainstruments.com/products/oem-probes.html) would help put to bed about 70% of the discussions here. This device provides 1% accuracy in air mass flow and 0.2% in repeatability. This slight modification would leave zero doubt about velocity profiles and turbulent flow, etc. since it measures mass flow not velocity. Also the turndown rate is 1000:1 and dynamic range is 0 to 20,000 SFPM so this device should allow replicators to measure a higher power level at higher temperatures accurately all the way up to 3000W.
If someone is successful replicating exactly as in the paper, if it were me, I would improve the air mass flow measurements and improve insulation with VIP technology and run the experiment again.
That panel has a max working temperature of +80C - a bit dangerous. I'd suggest anyone doing this stuff goes for materials that will withstand +150C at least to be safer, especially because better insulation => higher inner temperatures. Although airflow will cool this, the cooling will not be even and there will be hot spots.