Genes are often selected for molecular taxonomy citing their agreement to taxonomically derived species trees. These results are then used to define and edit existing taxonomic hierarchies of species. We counter the validity of this practice experimentally using mitochondrial genes routinely employed for the purpose. We constructed trees from complete mitochondrial DNA sequence (except D-loop) and its thirteen protein-coding genes namely ND1, ND2, ND3, ND4, ND4L, ND5, ND6, ATP6, ATP8, COI, COII, COIII, and CYTB of 255 species of Cypriniformes fish using the maximum likelihood and Bayesian methods. While ND4, ND5, COI, COII, COIII, and CYTB formed trees at par with the species tree, no gene produced complete concurrence. Similarly, while few genes formed similar branching patterns at family (COI and COIII) or broad group (ND1, ND2, and ND3; ND5 and COII) levels, no two gene trees fully agreed with each other. Tajima's D estimates of all studied genes indicated non-random evolution, violating the basic assumption of molecular phylogenetics. Bootstrap or Posterior probability values were also poor for several taxa indicating low robustness. Our findings reiterate that tree concurrence noted in studies involving smaller sample sizes is not a reflection of the evolutionary signature of the species.