The objective of this paper is to discuss the limitations in structural identification of large constructed structures. These limitations arise due to the geometric complexity, uncertain boundary and continuity conditions, loading environment, and the imperfect knowledge and errors in modeling such large constructed facilities. In this paper, the writers present their studies on developing a mixed microscopic-structural element level three-dimensional finite-element (FE) modeling of a long-span bridge structure and its structural system identification by integrating various experimental techniques. It is shown that a reasonable level of confidence (50-90%) can be achieved with a model that is calibrated using global and local structural monitoring data with a sufficiently high spatial resolution. The reliability of the global attributes, such as boundary and continuity conditions that may be identified and simulated by means of field-calibrated models using only dynamic test results (globally calibrated models), may appear to be high (as much as 90%). However, the reliability that should be expected for local stress fields is shown to be an entirely different matter, and a calibration based on just dynamic testing would be unable to reveal the confidence in simulated local responses. This is especially true for long-span bridges, because the resolutions of the dynamic test grids are often quite sparse due to the large size of the structures. In this paper, the writers illustrate that the density, modality, and bandwidth of experimental data should be carefully evaluated and matched to the size and complexity of a constructed system before claiming that a FE model is validated. It is also shown that even more than three dozen acceleration measurement points, two dozen strain measurements, and a continuous surveillance of wind and temperature were barely sufficient for a credible structural identification of a long-span bridge.