High-pressure partial melting studies are essential for understanding melt productivity, the influence of solid phases on melt composition, and the nature of the residue left in the deep Earth mantle. For instance, the variability of SiO2/MgO, CaO/Al2O3, and Al2O3/TiO2 ratios in komatiite compositions is generally attributed to the opx/ol ratios in the mantle source and the presence/absence of garnet during melting. Studies on mantle partial melting have been boosted by the development of activity-composition models for ultramafic and mafic systems, which are extensively utilised in thermodynamic modelling software. Although these models permit exploration of complex melting scenarios, the phase relations often diverge from experimental results, particularly at pressures >3 GPa. Here, we compare the phase relations obtained from partial melting experiments on a mildly depleted garnet lherzolite (GKR-001; 1400-1750ºC, 5GPa) conducted in a piston-cylinder apparatus with phase relations computed using MAGEMin (1400-1900ºC, 5GPa) for the same composition. At subsolidus conditions, the thermodynamic model fails to reproduce accurately the garnet and pyroxenes' experimental modes, overestimating garnet and orthopyroxene. The model and the experimental results agree with orthopyroxene at the solidus, but the solidus temperatures are >120ºC apart (GKR-001_exp = 1575ºC, GKR-001_calc = 1697ºC). Furthermore, the model narrowed the clinopyroxene stability field and expanded the orthopyroxene stability field to unrealistic melting degrees (96% melting). It also expanded the grt-out curve to higher degrees of melting than the experimental results (GKR-001_exp = 34%, GKR-001_calc = 50.6% melting). These discrepancies between experimental and thermodynamic results likely arise from the limitations of the thermodynamic activity-compositional models, uncertainties introduced by inconsistencies in the experimental data used to train these models, and choices regarding which experimental datasets should be included or excluded for model training. Uncertainties in the high-pressure peridotite solidus temperatures and phase relations restrict our understanding of geodynamics melt formation, especially on early Earth.