When I was at the old company, all the seismic analysis conflicted with observational data. The piping seismic engineers stuck with the analysis, because that was all they knew. Now, we have endless observations that peak ground velocity (PGV) correlates with damage, and never peak ground acceleration (PGA). Yet, they will stick with PGA until the cows come home to a destroyed barn. :)
They love giant shake tables, but that has nothing to do with reality. With a shake table or engineering analysis, the seismic force is represented essentially as a tilt, or lateral g force. In reality, a velocity pulse imposes a forced strain and the results are quite different. Here is a graphic of reality, and a design of a good testing machine.
The upper hand holds the top steady, while the bottom hand moves in an amount proportional to the PGV. To make a testing frame, you would replace the hands with clamps and impose a shear strain. If everybody had these machines, and threw out their shake tables, we would finally realize why soft soil amplification is a horrible thing. I can only hope.