Four published planet solutions fail the radius relation within three listed error widths and printed rounding

We tested the planet radius, planet-to-star radius ratio, and stellar radius in a NASA Exoplanet Archive PS snapshot from 22 October 2026. Selection for equal source references, complete errors, positive values, and valid flags leaves 5,143 rows for 3,882 planets. Six source rows fail the radius relation over full input ranges that span three listed error widths. Source checks confirm the values and errors in all six rows. Rounding in the printed central values reduces the list to five. Rounding in the printed error widths reduces it to four: GJ 3929 b, K2-43 b, K2-43 c, and TOI-2049 b. These four also fail when the nominal radius conversion varies by one percent. GJ 3929 b and TOI-2049 b still fail at five widths. The range test has no assigned probability. We also execute a hash-checked table helper from the released Hedges code. It prints a lower quantile as the central value and uses the wrong upper width for a median summary. We do not establish that this version made the final paper table or caused its radius conflicts. We release the exact archive response, source transcriptions, hashes, complete selection, bounds, numerical checks, figures, and reproduction code. This work extends prior public flags with error bounds and explicit rounding tests. It identifies solutions for review and does not give corrected planet radii or new planet detections.

NASA Exoplanet Archive · planet radius · data quality · transit geometry · interval bounds · printed precision · reproducibility