We consider a framework for linking two types of capture-recapture models (discrete-time and continuous-time models). Under our framework, the inference procedures for population size under the two types of models and the relative asymptotic efficiency are examined. It has been conjectured or intuitively thought that that the population size estimation under a continuous-time model is more efficient than that under a corresponding discrete-time model, but this conjecture/thought has not been theoretically justified in the literature. Our linking framework enables us to provide a theoretical justification for the MLE of population size when time or behavior response affect the capture probabilities. When both time and behavioral response exist, although a theoretical proof is currently not feasible, all numerical results also support the validity of the conjecture.