I am trying to validate a beam analysis against hand calcs, but I am getting differing results. So I set up a simple simulation to compare to have a base line.

Using weldments I modeled a 100" long S8 x 23 beam with a 25,000 lb point load in the center. Allowing the beams ends to rotate (simply supported) I get a very reasonable error of 3.3% on the deflection in the center, however with fixed ends the error is 21.3%. Why is the FEA computed deflection so much greater than the hand calcs?