Adenosine receptors, particularly A2BAR, are gaining attention for their role in pathological conditions such as cancer immunotherapy, prompting the exploration for promising therapeutic applications. Despite numerous selective A2BAR antagonists, the lack of selective full agonists makes the partial agonist BAY60-6583 one of the most interesting activators of this receptor. Recent cryo-EM structures have univocally revealed the binding mode of nonselective ribosidic agonists such as adenosine and its derivative NECA to A2BAR; however, two independent structures with BAY60-6583 show alternative binding orientations, raising the question of which is the physiologically relevant binding mode. In situations such as this, computational simulations that accurately predict shifts in binding free energy can complement experimental structures. Our study combines QligFEP and QresFEP protocols to directly compare the binding affinity of BAY60-6583 between alternative binding modes as well as providing a direct comparison of in silico mutagenesis studies on each pose with experimental mutational effects. Both methods converge on the experimentally determined binding mode that better explains both the existing SAR and mutagenesis data for this ligand. Our results allow the elucidation of the experimental binding orientation that should be considered as a basis for designing partial agonist derivatives with improved affinity and selectivity and underscore the value of free energy perturbation methods in aiding structure-based drug design.