Context : Recent work has used spectra of \sim 5000 stars in NGC 6397 that were extracted from a MUSE mosaic to determine the atmospheric parameters for these stars by fitting the spectra against the Göttingen Spectral Library .
A significant change in metallicity between the turn off and the red giant branch was found and was discussed as a possible manifestation of predicted effects of atomic diffusion .
However , the small amplitude of the effect and inconsistency with earlier measurements call for more attention before this result is interpreted .
Systematic effects due to the interpolation or to the synthetic spectra can not be ruled out at this level of precision .

Aims : We reanalyze the data with : the ELODIE and MILES reference libraries in order to assess the robustness of the result .
These empirical libraries have a finer metallicity coverage down to approximately the cluster metalicity turn-off .

Methods : We used the ULySS full-spectrum fitting package , together with the library interpolators to remeasure the three atmospheric parameters effective temperature , surface gravity , and [ Fe/H ] \xspace metallicity .

Results : We find a very low [ Fe/H ] \xspace dispersion along the isochrone ( 0.07 dex ) , consistent with our error estimate ( 0.05 dex ) .
However , the [ Fe/H ] \xspace trend is not reproducible .
This shows that the data have the potential to reveal patterns of the magnitude of the expected physical effects , but the analysis methods need to be refined to cull systematic effects that currently dominate the patterns .

Conclusions :