Job scheduling in manufacturing companies faces potential challenges due to varying job durations per machine, working calendar structures, and preventive maintenance. This must be considered to avoid unplanned outages and respect planned downtime. This research proposes and analyzes three mathematical models to solve the Unrelated Parallel Machine scheduling problem (UPM), considering machine availability and eligibility constraints. The developed models address the optimization of two major metrics: the makespan, and the completion time of the last shift used. Minimizing these measurements is crucial to maximize system efficiency and reduce high-demand industries' operating costs. The models differ in their structure and how they were conceptualized. Model 1 uses sequence-dependent variables, offering detailed job sequencing but increasing problem size as job counts rise. Model 2 simplifies by assigning jobs to machines and shifts without sequencing, improving computational efficiency but compromising detail. Model 3 balances detail and efficiency by incorporating start-time-dependent variables, suitable for large-scale problems. Tests were performed varying the number of jobs, machines, shift lengths, and mean job durations. The results show that Models 2 and 3 outperform at makespan minimization, obtaining an optimal solution to almost all benchmark instances within an hour. Model 1 struggles with scalability, achieving less than 0.5% GAP in only 69% of small cases and failing for larger setups. For the cases where the minimization of the completion time of the last operating shift is considered, all models improve computational times and GAPs. These findings guide the selection of scheduling models based on available computational resources and detailed planning needs in high-demand industrial environments.