Designing genetic networks with desired functionalities requires an accurate mathematical framework that accounts for the essential mechanistic details of the system. Here, we formulate a time-delay model of protein translation and mRNA degradation by systematically reducing a detailed mechanistic model that explicitly accounts for the ribosomal dynamics and the cleaving of mRNA by endonucleases. We exploit various technical and conceptual advantages that our time-delay model offers over the mechanistic model to probe the behavior of a self-repressing gene over wide regions of parameter space. We show that a heuristic time-delay model of protein synthesis of a commonly used form yields a notably different prediction for the parameter region where sustained oscillations occur. This suggests that such heuristics can lead to erroneous results. The functional forms that arise from our systematic reduction can be used for every system that involves transcription and translation and they could replace the commonly used heuristic time-delay models for these processes. The results from our analysis have important implications for the design of synthetic gene networks and stress that such design must be guided by a combination of heuristic models and mechanistic models that include all relevant details of the process.