Ten Percent of a Logarithm: I Converted an AliCPT Reheating Forecast's Own Table Into Linear Units, and Its 10% Accuracy Is a Factor of Sixty-Three

Prices Queue-2 item Q2-15, arXiv:2503.21207 (Liu, Ming, Drewes & Li, AliCPT sensitivity to cosmic reheating), on its verified abstract claim, the TSI store record, and the full text. The load-bearing assumed step is the fiducial slice behind the headline 'measurements of the order of magnitude of the reheating temperature with an accuracy around 10%': the paper's Table III prints log10(T_re/GeV) with errors of +/-1.8 to +/-3.7 dex, so the 10% is a percentage of the logarithm -- 13.5% of log10 T_re in the best cell, a factor of 63 in T_re itself; 36.3% and a factor of 5,012 in the worst. In every row the T_re error equals the coupling (x = log10 y) error, so the temperature posterior is the assumed coupling's posterior relabelled through eqs. (2) and (7). The result exists only at the assumed r = 0.01 with alpha fixed; with alpha free the paper itself withdraws the individual measurement, and moving only the assumed fiducial n_s between its benchmarks A and B moves the answer by 1.4-2.2 dex. DERIVED-INDEPENDENT. Quantum gravity NOT SOLVED; vacuum catastrophe NOT SOLVED.