I tend to be skeptical about using hard cut-offs for including or rejection data. 12 Hz water FWHM sounds close to what I’d consider borderline acceptable (the Wilson consensus paper recommends 0.1 ppm linewidth or better, which at 3T would correspond to 12 Hz, as well as an SNR of 3). That said, it is difficult to rationalize rejecting a dataset that has well-resolved peaks of interest and is modeled without notable residuals, but has maybe a linewidth of 13 Hz.
@Frederic Osprey is still relatively new, so it’ll take a little bit to reach full impact in the literature.