This work proposes bounded instances for the Job Shop Scheduling Problem that we tailored using Unified Particle Swarm Optimization. We developed and evaluated 2000 specialized instances across five sizes to evaluate heuristic performance under controlled, unbiased conditions. Our primary objective is to mitigate the inherent biases of random instances that favor specific heuristics. Our three-phase methodology begins by determining a suitable delta value, generating instances, and analyzing their features. The results indicate compliance with the delta values and balanced heuristic performance. This framework enhances the fairness of algorithm benchmarking and provides a basis for research in solver selection, metaheuristics, and adaptive learning in combinatorial optimization.