Evidence for the 340-km-long Fort Tejon earthquake of 1857 is found at each of the high-resolution paleoseismic sites on the southern San Andreas Fault. Using trenching data from these sites, we find that the assemblage of dated paleoearthquakes recurs quasi-periodically (coefficient of variation, COV, of 0.6, Biasi, 2013) and requires ~80% of ruptures were shorter than the 1857 rupture with an average of Mw7.5. In contrast, paleorupture lengths reconstructed from preserved geomorphic offsets extracted from lidar are longer and have repeating displacements that are quite regular (COV=0.2; Zielke et al., 2015). Direct comparison shows that paleoruptures determined from geomorphic offset populations cannot be reconciled with dated paleoearthquakes. Our study concludes that the 1857 rupture was larger than average, average displacements must be < 5 m, and suggests that fault geometry may play a role in fault behavior.