Classic morphological studies of the oldest, so-called nonditrysian lineages of Lepidoptera yielded a well-resolved phylogeny, supported by the stepwise origin of the traits characterizing the clade Ditrysia, which contains over 98% of extant lepidopterans. Subsequent polymerase chain reaction (PCR)-based molecular studies have robustly supported many aspects of the morphological hypothesis and strongly contradicted others, while leaving some relationships unsettled. Here we bring the greatly expanded gene sampling of RNA-Seq to bear on nonditrysian phylogeny, especially those aspects that were not conclusively resolved by the combination of morphology and previous PCR-based multi-gene studies. We analysed up to 2212 genes in each of 28 species representing all 12 superfamilies and 15 of 21 families of nonditrysians, plus trichopteran outgroups and representative Ditrysia. Our maximum likelihood phylogeny estimates used both nonsynonymous changes only (degen1 coding) and all nucleotides (nt123) partitioned by codon position, recovering a novel hypothesis for early glossatan relationships that is the most strongly supported to date. We find strong support for Micropterigidae alone as the sister group to all other Lepidoptera, in agreement with morphology and early molecular evidence, but in contrast to recent PCR-based studies. Also very strongly supported are the previously recognized clades Angiospermivora, Heteroneura, Eulepidoptera and Euheteroneura. Finally, we find strong support for paraphyly of the southern hemisphere family Palaephatidae, with the South American genus Palaephatus Butler forming the previously undetermined sister group to Ditrysia. The remaining palaephatids, Australian and South American, form the sister group to Tischeriidae.