Optical biosensors provide a platform for qualitatively and quantitatively analyzing various biomolecular interactions. In addition to advantages such as label-free and high-throughput detection, these devices are also capable of measuring real-time binding curves in response to changes in optical properties of biomolecules. These kinetic data may be fitted to models to extract binding affinities such as association rates, dissociation rates, and equilibrium dissociation constants. In these biosensors, one of the binding pair is usually immobilized on a solid substrate for capturing the other. Due to the nature of these surface-based methods, mass transport effects and immobilization heterogenetity may cause problems when fitting the kinetic curves with the simple one-to-one Langmuir model. Here real-time binding curves of various antibody-antigen reactions were obtained by using an ellipsometry-based biosensor, and the results were fitted to the simple one-to-one model as well as a more sophisticated approach. The results show that the one-to-two model fitted much better to the curves than the one-to-one model. The two-site model may be explained by assuming two immobilization configurations on the surface. In summary, in fitting real-time curves obtained from optical biosensors, more sophisticated models are usually required to take surface-related issues, such as immobilization heterogenetity and mass transport effects within targets, into account.