The Unified Modeling Language (UML) remains a core topic in Software Engineering education and is traditionally introduced in courses on software analysis, design, and software architecture. For decades, UML has been regarded as a fundamental instrument for teaching abstraction, system decomposition, and reasoning about structure and behavior in complex software systems. However, the increasing adoption of agile development practices and lightweight documentation has challenged traditional approaches to teaching software modeling, raising questions about what should be taught and how strongly UML-based instruction should be emphasized in modern curricula. From a Software Engineering education perspective, understanding how UML is taught, learned, and used in professional contexts is essential for aligning academic curricula with industry practices while preserving fundamental design competencies. Educators face the challenge of maintaining conceptual rigor in modeling education while preparing students for professional environments in which modeling is often selective and shaped by agile workflows. This paper reports a cross-national empirical study based on two large-scale surveys conducted in Brazil and Portugal, involving 80 higher-education instructors and 206 students and software practitioners. A mixed-methods research design was employed, combining descriptive statistical analysis with qualitative analysis guided by Grounded Theory. The study investigates curricular placement of UML, instructional strategies, diagram selection, tool usage, learning challenges, and perceived relevance to professional software development. The results show that UML instruction in higher education concentrates on a limited set of diagrams, mainly class, use case, and sequence diagrams, and is typically embedded within traditional Software Engineering courses. In contrast, students and practitioners report selective and infrequent use of UML in industry, especially in agile development contexts. Key educational challenges include limited integration of modeling activities with authentic development tasks, insufficient exposure to modeling tools, and misalignment between academic expectations and professional practice. The study provides clear value for Software Engineering education by offering evidence-based recommendations for curriculum design and teaching practice, including agile-compatible modeling instruction, stronger integration with practical development activities, and positioning UML as a pedagogical instrument for abstraction, communication, and design reasoning rather than exhaustive documentation.