A single-flow model fails to account for localized low-z anisotropy in the Hubble residual field

We test whether the low-redshift Hubble diagram on the Pantheon+SH0ES footprint can be compressed into a scalar expansion rate plus a single flow correction. The residual field is not exhausted by the standard kinematic dipole with $1/z$ scaling: beyond that term, the data require a second directional component whose tomographic signature is localized around $z\simeq 0.027$ and forms a rise-peak-extinction pattern across adjacent redshift shells. We confront this morphology with an independent CF4++ velocity-field reconstruction and find that the external field captures part of the broad low-$z$ flow-compatible signal, but not the shell-localized TEX structure. The result is therefore structured rather than merely residual: Pantheon isolates a localized second directional component, while CF4++ recovers only the broader flow-compatible regime and fails on that localized feature. This Pantheon-CF4++ mismatch is the main empirical result. A dedicated shell-level null calibration then shows that reproducing simultaneously the observed shell peak, adjacent-shell extinction, weak CF4 contribution, and CF4 non-absorption under $\mathcal{M}_0:\Delta\mu=a_0+\beta_{\rm CF4}t_{\rm CF4}+\epsilon$ on $0.020<z\le0.040$ has empirical probability $p_{\rm joint}\approx 2.0\times10^{-6}$. We therefore find a calibrated failure of the standard local scalar-$H_0$ plus single-flow description on the observed low-$z$ footprint.