Ensuring objectivity in expert-based educational program evaluations remains challenging due to inherent subjectivity and uncertainty of the assessment processes. In this study, we apply fuzzy signature models to address these issues within the evaluation framework of the Hungarian Accreditation Committee (MAB). A generalized fuzzy signature structure was developed, that supports discrete answer categories and hierarchical question groups. We systematically investigated how key parameters-fuzzy number shape, aggregation operators, and defuzzification methods-affect final evaluation outcomes. Two types of fuzzy coverage sets (core-based and support-based) were tested across multiple configurations using simulated data (k=100,000) aligned with typical expert response distributions. Statistical analysis, including paired t-tests and normality tests, revealed that the Largest of Maximum (LoM) defuzzification method combined with arithmetic mean aggregation and a core-to-support ratio of 0.75 yielded the highest classification accuracy (83.77%) and the most stable results. Core-based (Type 1) coverage consistently outperformed support-based (Type 2) coverage in 98% of tested configurations. These findings provide practical guidance for the design of robust and interpretable fuzzy evaluation systems and support the development of more transparent quality assurance tools in higher education. This work goes beyond mere comparison by systematically analyzing the sensitivity of fuzzy evaluation systems to parameter choices, which is critical for robust system design of accreditation models.