PREMISE OF THE STUDY: Reconstruction of plant phylogeny has heavily relied on single-gene or multigene plastid data. New sequencing methods have led to an increasing number of studies based on data from the entire plastid, but the mitochondrion has rarely been used to infer plant phylogeny because of an assumed information poverty and demonstrated lateral transfer of mitochondrial gene regions between distantly related species. METHODS: We explored phylogenetic information from the plant mitochondrion using 57 representatives of the species-rich coffee family as study system and assessed consistency with previous results based (mostly) on plastid data. KEY RESULTS: We showed that the mitochondrial genome can provide structured and statistically significant information on plant phylogeny. While most of our results are consistent with those based on plastid data, some surprising and statistically significant conflicts emerge, and our study demonstrates with striking clarity that the phylogeny of Rubiaceae is far from resolved. CONCLUSIONS: It appears unlikely that conflicts between results retrieved from the different genomic compartments would be restricted to Rubiaceae. Rather, they are probably a general phenomenon and an important factor behind longstanding difficult phylogenetic questions. The biological processes responsible for the conflicting results detected here are unclear, but some conflicts are likely caused by hybridization events that occurred tens of millions of years ago. Whether such ancient events can be reconstructed based on molecular data from extant plants remains to be seen, but future studies of the nuclear genome may provide a way forward.