Remark: What is corrected, and what is not

The predecessor's proposed caption change from $+\mathrm{i}{u}^{(0)}$ to $-\mathrm{i}{u}^{(0)}$ follows algebraically from $v=u/\mathrm{i}$, since $1/\mathrm{i}=-\mathrm{i}$. It remains a documentary normalization observation. It cannot make $a^{(0)}=-f^{\mathrm{reg}}$ true.

The typeset and regression values are both retained. Their physical origin cannot be decided from this comparison: it does not identify which underlying loop artifact, if any, contains the correct coefficient. In particular, it is inaccurate to describe the witness as a failure of every entry in the typeset gauge table. The failure is specifically cross-artifact.