The period and the blue amplitude of the light curve are used to calculate the light-to-mass ratio A of ab-type RR Lyrae stars in a wide sample of Galactic globular clusters. It is shown that the resulting (reddening-free) mean light-to-mass ratio depends on the horizontal-branch (HB) morphology as well as the metallicity of the cluster, in full agreement with the prescriptions of synthetic horizontal-branch computations. As a consequence, also the slope of the observed relationship between A and [Fe/H] turns out to vary according to the HB morphology of the clusters taken into account. Specifically, if clusters with not very blue HB are taken into consideration, then the relationship has a slope DELTAA/DELTA[Fe/H] approximately -0.05, which is much milder than that (approximately -0.10) obtained by Sandage from the analysis of the period and effective temperature of ab-type RR Lyrae stars in the field and in a selected sample of clusters with not very blue HB. However, the inclusion of clusters with very blue horizontal branch yields a steeper slope (DELTAA/DELTA[Fe/H] approximately -0.08), which is virtually the same as that found for the field variables. Such observational results would suggest some correction to the reddenings and/or color-temperature calibration adopted by Sandage. Furthermore, the above results would confirm the suggestion of Lee that many metal-poor variables in the field are highly evolved stars from the blue side of the RR Lyrae gap. Finally, it is shown that all the observed light-to-mass ratios could be fitted with a constant primordial helium abundance Y0 = 0.23, which is the value suggested by the ratio R of the total number of HB stars to the number of giant stars brighter than RR Lyrae variables.