Disentangling Likes and Dislikes in Personalized Generative Explainable Recommendation