# Limitations

- Published component-removed curves are posterior spectral post-processing,
  not independently optimized null retrievals.
- Spectral covariance matrices and posterior samples are not present in Source
  Data Fig. 3; calculations use diagonal observational errors.
- All binning designs reuse the same observations and are sensitivity analyses,
  not independent replications or a multiple-testing route to significance.
- Linear interpolation of model contours is adequate for this display-scale
  diagnostic but is not an instrument line-spread-function integration.
- The simple abundance comparison symmetrizes asymmetric marginal intervals
  and assumes independence between transmission and emission posteriors.
- The scale height uses a fixed mean molecular weight and equilibrium
  temperature; it is illustrative, not a retrieved atmospheric altitude.
- Only TESS Sector 54 is available in the queried SPOC time-series result. The
  transit fit fixes limb darkening, scaled semi-major axis, eccentricity, and
  dilution assumptions and is not a global transit-timing solution.