Personalized insulin dosing for Type 1 diabetes mellitus (T1DM) remains challenging because of complex glucose–insulin dynamics and substantial patient variability. Reinforcement learning (RL) has emerged as a promising approach for adaptive insulin management, yet the reliability of learned policies depends heavily on reward design and evaluation strategy. This study compares three actor–critic RL algorithms: Soft Actor-Critic (SAC), Advantage Actor-Critic (A2C), and Proximal Policy Optimization (PPO) for personalized insulin dosing using real-world continuous glucose monitoring, insulin delivery, basal insulin, and meal intake data from the OhioT1DM dataset. A custom Gymnasium-based environment was developed, and all algorithms were trained under identical conditions for 100,000 timesteps. Performance was evaluated using cumulative reward together with clinically relevant measures, including Time in Range (TIR) and insulin dosing behaviour. Although A2C and PPO achieved higher cumulative rewards than SAC, both converged to near-zero insulin dosing policies that exploited the reward formulation rather than learning clinically meaningful glucose regulation. In contrast, SAC maintained adaptive dosing behaviour, achieving a TIR of 72.71% with an average insulin dose of 1.769 U/step. These findings show that higher cumulative reward does not necessarily correspond to better clinical decision-making in open-loop reinforcement learning environments. The study highlights the importance of behaviour-focused evaluation alongside conventional reward metrics and provides practical insights for developing safer and more reliable reinforcement learning systems for personalized diabetes management.