Terminology summary
Reference ligand for bias: The ligand that is, by definition, unbiased. The bias of other tested ligands are quantified relative to this reference.
Reference ligand for Emax: If the reference ligand for bias does not produce a full receptor response, a separate reference ligand for Emax can be used to exploit the full window of response. For the Gs pathway, some studies use forskolin, which activates adenylate cyclase directly, rather than a reference ligand to determine the maximal response. Both the reference compound for bias and the ligand tested for bias will have their Emax values assigned relative to the reference ligand(s) for Emax for each pathway.

Biased ligands/signaling (using any reference ligand)

When using any ligand (e.g. a tool compound or clinically relevant ligand), the statement that a tested ligand is biased can only be made relative to the reference ligand (which itself can have any bias).
Recommendation 2: Our recommendation here is to report clearly the identity of the reference compound and to restrict the conclusions to the comparison made. To facilitate cross-study comparison, it is also recommended to consider if a reference ligand can be chosen that has been used repeatedly for the given receptor and pathways before. Commonly used reference ligands are available in the receptor pages in the Guide to Pharmacology database (Armstrong et al., 2020).
Caveat: When the reference ligand is not chosen based on its signaling bias and is not the principal endogenous agonist, the results cannot be used to determine pathway- or physiology-bias (see Table 1 and the two next sections).