Linear Contextual Bandits with Hybrid Payoff: Revisited