This paper presents an in-depth investigation of the Power Measure (PM) applied with varying values of the exponent q in Choquet-based generalizations for fuzzy rule-based classification systems (FRBCS). We evaluate ten fixed values of q in the interval [0.1, 1.0] across 33 datasets, assessing classification performance through non-parametric statistical analyses. Among the generalizations, the \(C_{F1F2}\) -integral consistently achieves superior results, reaffirming its state-of-the-art status within this domain. Our findings reveal that optimal q values are inherently method-dependent: \(q=0.4\) is best suited for Choquet and CC-min, \(q=0.6\) for CT, \(q=0.2\) for CF \(_{\text {Avg}}\) and \(C_{F1F2}\) , \(q=0.8\) for dCF, and \(q=0.5\) for dXC. Statistical comparisons using the Aligned Friedman and Wilcoxon signed-rank tests show that while PM performs competitively across most configurations, it exhibits occasional underperformance when compared to specific generalizations such as Choquet and dCF. The results helps in the comprehension of method-specific q optimization and provide practical guidance for selecting aggregation strategies in FRBCS.