Atomic diffusion (AD) is a slow continuous process that depletes heavy elements such as Fe from the surface layers during the MS lifetime of a star. As the star evolves to the RGB the effect of depletion disappears as its deep outer convection zone restores the original composition in the atmosphere. Mixing processes at work below the outer convection zone reduce the effect of AD as they hinder the downward diffusion of heavy elements, and thus a more efficient mixing will cause flattened diffusion trends. Such additional mixing (AddMix) seems to be needed to reproduce observed abundance trends. Although the inclusion of extra mixing in a layer just below the convective envelope is remarkably successful in describing the observed abundance trends in NGC 6397 ([Fe/H] = −2.1), the description applied is in no way unique or physically satisfying. To better understand the physics and place additional constraints on the possible variation of extra mixing with stellar parameters such as metallicity, we conducted a study, similar to that presented in Korn et al. (2007), of another metal-poor GC NGC 6752 ([Fe/H] = −1.6). In Fig. 1 we show the results, published in Gruyters et al. (2013), which shows weak yet systematic abundance trends with evolutionary phase for Fe, Sc, Ti and Ca. The trends are best explained by stellar structure models including AD with efficient AddMix. As a consequence sub-primordial stellar lithium abundances of the stars on the Spite plateau can be brought up to match the WMAP-calibrated Big Bang nucleosynthesis predictions to within the mutual 1-sigma errors.