Phylogenomic datasets often consist of reduced representations of the genome, but differences inherent to the nature of these genomic markers may affect their performance. The information content of individual genomic markers is rarely interrogated, let alone compared to different types of markers from target-capture sequencing methods. This information will be important for deciding which data types are better suited for resolving the relationships of interest. We explored the causes of phylogenetic disagreement in the relationships of heroine cichlids by comparing topologies, support, phylogenetic informativeness, gene tree heterogeneity, and other sources of discordance based on ultraconserved elements and exons. We found that both types of genomic markers are affected by low phylogenetic signal-to-noise ratios, which for ultraconserved elements might be driven by incomplete lineage sorting and for exons by systematic errors from overly generalized model-fitting. More accurate modeling of substitution rates from individual loci, and of dealing with ILS, will likely increase the accuracy of phylogenomic trees and resolve the discordance between datatypes in the future.