Ontologies are commonly used in computer science either as a reference model to support semantic interoperability, or as an artifact that should be efficiently represented to support tractable automated reasoning. This duality poses a tradeoff between expressivity and computational tractability that should be addressed in different phases of an ontology engineering process. The inadequate choice of a modeling language, disregarding the goal of each ontology engineering phase, can lead to serious problems in the deployment of the resulting model. This article discusses these issues by making use of an industrial case study in the domain of Oil and Gas. The authors make the differences between two different representations in this domain explicit, and highlight a number of concepts and ideas that were implicit in an original OWL-DL model and that became explicit by applying the methodological directives underlying an ontologically well-founded modeling language.