Intermolecular interactions are difficult to model, especially in systems formed by multiple interactions. Such is the case of caffeine-phenol. We extracted structural data using mass-resolved excitation spectroscopy and double resonance techniques. Then we explored the predictions of seven different computational methods, to discover structural and energetic discrepancies between them that may even result in different assignments of the system. The results presented here, highlight the difficulty of constructing functionals to model systems with several competing interactions, and rise the alarm about assignments of complex systems with limited experimental information, relying exclusively on energetic data.