The phylogeny of the class Actinobacteria remains controversial, essentially because it is very sensitive to the choice of dataset and phylogenetic methods. We used a test proposed recently, based on complete genome data, which chooses among candidate species phylogenies based on the number of lateral gene transfers (LGT) needed to explain the diversity of histories among gene trees for a set of genomes. We used 100 completely sequenced genomes representing 35 families and 17 orders of the class Actinobacteria and evaluated eight different hypotheses for their phylogeny, including one based on a concatenate of 54 conserved proteins present in single copy in all these genomes, trees based on 16S and 23S rRNA gene sequences or their concatenation, and a tree based on the concatenation of MLSA genes (encoding AtpI, GyrA, FtsZ, SecA and DnaK). We used Prunier to infer the number of LGT in 579 proteins (different from those used to build the concatenated tree) present in at least 70 species, using the different hypothetical species trees as references. The best tree, with the lowest number of lateral transfers, was the one based on the concatenation of 54 proteins. In that tree, the orders Bifidobacteriales, Coriobacteriales, 'Corynebacteriales', 'Micromonosporales', 'Propionibacteriales', 'Pseudonocardiales', Streptomycetales and 'Streptosporangiales' were recovered while the orders 'Frankiales' and Micrococcales were not. It is thus proposed that the order 'Frankiales', which has an effectively but not validly published name, be split into Frankiales ord. nov. (type family Frankiaceae), Geodermatophilales ord. nov. (Geodermatophilaceae), Acidothermales ord. nov. (Acidothermaceae) and Nakamurellales ord. nov. (Nakamurellaceae). The order Micrococcales should also be split into Micrococcales (genera Kocuria, Rothia, Micrococcus, Arthrobacter, Tropheryma, Microbacterium, Leifsonia and Clavibacter), Cellulomonales (Beutenbergia, Cellulomonas, Xylanimonas, Jonesia and Sanguibacter) and Brachybacteriales (Brachybacterium) but the formal proposal for this will have to wait until more genomes become available for a significant proportion of strains in this order.