Anticorrelations of p-capture elements in the GC NGC 2808 (Carretta 2015). The diagram of sodium-magnesium data (panel b) was the basis for the subdivision into different sub-classes: P1 (blue), P2 (green), I1 (red), I2 (orange) and E (black). All panels show: (1) standard average abundances in the AGB ejecta by for a metallicity Z = 2.3 × 10⁻³, [α/Fe] = 0.4, e.g. [Fe/H] = –1.08 (open blue circles); (2) exploratory models (coloured squares and connecting lines) with the same initial composition and identified by the following. GREEN: Ejecta of the 6 M_⊙, with standard neon and decreasing mass-loss rates (models ηxNe1). Starting from the standard abundances for η = 0.02, we plot η = 0.01 (1/2 standard), 0.005 (1/4 standard), and 0.002 (1/10 standard); BLUE: 6 M_⊙ for η = 1/10 standard, with increasing neon abundance from standard (solar scaled, for α/Fe = 0.4) to two and four times standard (models η 0.1Ne4); RED: Abundances for the 5 and 7 M_⊙ for η 0.1Ne4; CYAN: Starting from mass loss 1/4 standard, and increasing neon at two and four times the standard abundance (models η 0.25Ne2 and η 0.25Ne4). The big blue and cyan stars define the location of the models that have sodium consistent with the O-Na patterns and discussed in the text: η 0.1Ne4 and η 0.25Ne2, which we will also consider in the other panels. The dark grey lines with triangles define dilution of the 6 M_⊙ location η 0.25Ne4 with increasing quantities (20, 40, 60 and 80%) of gas with standard composition, identified by the yellow star.