During the design phase of a wind turbine, it is common practice to decouple global load simulations from detailed structural analysis to reduce computational cost. This is achieved by constructing a global model of the turbine system from low- to mid-fidelity sub-models, under the assumption that nonlinear structural effects and localized modeling inaccuracies will have only a limited effect on the overall system dynamics. However, the considerable number of unexpected component failures, particularly offshore, have raised the question of whether such assumptions remain valid across the full operational envelope. In addition, as turbine sizes continue to increase, nonlinear structural effects may become more important, especially in rotor blades of floating systems, to an extent that these assumptions may no longer hold. In response, many simulation tools now support higher-fidelity blade representations, such as geometrically exact beam models. This study presents an alternate higher-fidelity version of the OpenFAST global model of the 15 MW IEA reference floating offshore wind turbine, in which the blades are represented as shell elements in Abaqus. The simulation framework relies on staged, bidirectional coupling between Abaqus and OpenFAST. To build confidence in the model, we first verify its individual components: the shell blade model itself, the aerodynamic load transfer method (which converts point loads from AeroDyn into distributed loads for the shell mesh), and the conversion of shell deformations into an equivalent beam-based representation. Finally, we perform a code-to-code comparison between the fully coupled shell-based model and a standard BeamDyn-based OpenFAST simulation. Preliminary results show strong agreement between the two approaches under normal operating conditions, supporting the validity of the model for intact blade conditions (i.e., as manufactured, without in-service degradation). However, under more extreme loading scenarios, some differences begin to emerge, highlighting the potential limitations of beam models in accurately capturing blade-level dynamics and structural responses. These differences may increase in applications that account for degradation-driven changes in blade mechanical properties, with more pronounced local phenomena and cross-sectional nonlinearities.