For this example, the FEA model with true properties and diameters will not give the same forces as Shigley.

Shigley is still useful because there are many ways to make mistakes while building an Ansys model, so knowing what the correct answer is within 15% by a hand calculation allows you to find mistakes that result in a 100% or 1000% error in the Ansys model.

FEA models are validated against experimental data. Once you have a physical sample of a bolted joint, it is easy to apply a known load (most easily done by hanging known weights) and carefully measuring deflection. A discrepancy of 20% is a good result unless extraordinary care is taken to characterize the stiffness of the I-beam. One method is to extend the model past that cut plane and add the full length of the I-beam and the details of how that is bolted to a concrete foundation. With the Fixed Support on a close cut boundary, the deformation from the Ansys model will underpredict the actual tip deflection by a lot.