Any computer model needs to be verified for errors by comparison with hand calculation, test data. Think about cost/benefit ratio, risk assessment processes and how difficult things would become without such model verification?
It is often a trade-off between scale and accuracy as far as a CFD/FEA model is concerned. An overview of the airflow can be obtained through a coarse model, but for really accurate results this model must be simplified to specific areas with certain assumptions made. You can make many smaller, more accurate, models by taking the results of each calculation, insert them as boundary conditions in the model for the next bit of the structure and so forth to enable assessment of the interaction downstream of aircraft components.
However, how do you assess for small errors in the calculation? These would be passed along the model from calculation to calculation.
My answer to your question: it depends on who's funding the project and how accurate they want their results to be...
