As key component of the shield tunnel, the longitudinal joint will directly affect the mechanical behavior of the tunnel structure. When considering structural effect of the joint under compression-bending scenario, existing researches always assume that the joint surface remains flat when the joint rotates, which is also the basic assumption of the rigid plate spring (RPS) method. However, the assumption and the RPS method have some limitations in matching with structural analysis, mainly because of the empirical value of calculation length and the inaccurate description of the joint deformation mode. In order to reveal the joint bending mechanism, distribution of contact pressure, segment stress, and segment deformation of the joint should be firstly investigated. Thus, an elastoplastic modified analytical model based on a 2D flat rectangular plate is established. It is found that the contact pressure distribution is not linear, but a saddle-shaped distribution at the beginning. With the increase of the bending moment, the detached zone and the triaxial compression zone both appear and develop. It is found that there is a transition zone called joint influence zone, within a certain distance from the joint surface, where the distribution of segment stress and deformation transitions from nonlinear to linear. The real joint rotational angle is also defined, which comes from two parts, including contact flaw and geometric additional stress, where the contact flaw is the reason why initial joint stiffness increases as the axial force increases. By comparing calculation results with the test results, the applicability of the model is validated, and measurement limitations of the real joint tests are also revealed. Some suggestions are proposed for several simplified models, considering structural effect of the longitudinal joint in tunnel structure, including stiffness reduction beam model, beam-spring model and stiffness-reduced beam-spring model.