My experience from the group I have been working with (Drosophilidae) is that molecular data is far superior to the morphological data. Morphological data is always under some form of selection, while big chunks of molecular data are under far less selective pressures and therefore more likely to reflect phylogenetic history. In perfect situations, both match up nicely, but when not, you have to choose which sets of traits are more likely to provide you with the correct answer.
It is possible to obtain modern molecular data, but difficult beyond 200-2000 years ago, when the molecules are too degraded, even if the samples are available. Prior to that we must rely on morphology mostly. Molecular data tends to have many more characters, so can potentially give much higher resolution, and possibly less subject to technical problems such as convergent evolution, but not entirely. Molecular characters are getting cheaper to obtain each year. Perhaps the best answer is to use Bayesian methods which can integrate both molecular and phenotypic data.
I would prefer to analyse my morphological data (descriptively) before going into molecular but I did not do any phylogenetic analysis yet. Sometimes molecular could be misleading when you wrongly identify such species or misidentified or analysing a cryptic species. However everything is depending on your research question. Phenotypic data can also be confusing if the taxa you are working is prone to environmental changes. I hope this would give you some idea.