Years passed from first time using the finite element method to simulate almost all cases within a recommendation that its fit to predict a similar behavior of the experimental programs, .. What did you think upon your opinion...?
Finite element methods are powerful tool in analysis and design airplanes, space rockets,submarines etc.,but in structural Eng. the strategy of adapting such methods to replace the experimental work didn't followed yet.
In many civil engineering problem the complexity of such problem let the analytical solution or experimental modelling is difficult in simulating. Hence, the numerical analysis with assumption make the solution approximately but acceptable. Problems of cracks in concrete, jointed rock masses, flow in joints and anisotropic material etc.
it is hard to conclude that FEM can replace the experiment result perfectly.however approximate result can be obtained.now days there are a lot of FE software but the capability of all software is different as Moataz A. said that there are some responses that are difficult to be captured using FEM.the software available now days ,some of them are based on the quantum model,other are based on the fiber model,and other are base on concentrated hinge (this is important if you are dealing with nonlinear analysis)so depending on the analysis you are doing you need to make verification on the FE software then compare the result,if the response found from experimental captured by the software,it means that the FE software is okay,for the specific analysis type you compared.