“…The quality of the atomic models used for calculation of omit ED maps does affect the accuracy of the absolute values of ENs determined, more so for the absolute ENs than for the relative ENs. For small molecules, model R -factors of crystal structures are refined to 1–2% at subangstrom resolution. ,,, However, model R -factors for protein structures are much larger. For example, the cross-validation free R -factor for the 6jlj structure is 19.8% at 2.15 Å resolution, while for the 6w1o structure, it is 24.0% at 2.08 Å resolution. , Therefore, the results obtained for the 6jlj structure should be more accurate and more reliable than those for 6w1o. , Here, we use one of the metal ions (Mn2 of each OEC) as an internal reference, which improves greatly the accuracy and reliability of relative ENs because many modeling errors tend to cancel out (see the Supporting Information).…”