The hyperdiverse genus Sarcophaga Meigen, with about 890 valid species arranged within 169 subgenera, accounts for almost half of the diversity of the subfamily Sarcophaginae. Current phylogenetic hypotheses for this genus are poorly supported or based on small taxon sets, or both. Here, we use molecular data from the genes COI and 28S to reconstruct the phylogeny of Sarcophaga based on the most comprehensive sampling for the group to date: 144 species from 47 subgenera, including representatives from all regional faunas for the first time. Of the total sequences of Sarcophaga used in the present study, 94.7% were newly generated. The secondary structure of the D1–D3 expansion segments of 28S is presented for the first time for the family Sarcophagidae, and is used in a multiple sequence alignment. Branch support and tree resolution increased remarkably through rogue taxa identification and exclusion. Rogue behaviour was explained mostly as a missing data problem. The RogueNaRok web service and the algorithms chkmoves, IterPCR and prunmajor implemented in the computer program TNT were equally good at identifying critical rogue species, but chkmoves and IterPCR also identified rogue clades. Pruning rogues increased the number of monophyletic subgenera in consensus trees from one to six out of 19 subgenera with more than one representative species. Bayesian inference, maximum‐likelihood and parsimony analyses recovered more monophyletic subgenera after the removal of rogue taxa, with parsimony showing the largest improvements in branch support and resolution. Although with low support, Nearctic taxa were found to be the earliest diverging lineages, followed by a subsequent diversification of Old World faunas, which is in agreement with currently available evidence of a New World origin and early diversification of Sarcophaga.