Chris,

As I recall, their Science paper showed that the motions would be much worse for a realistic basin structure than for a structure with no basin. However, the engineers already include such site effects in their estimates of ground response in basins. The data close in to past large earthquakes was very sparse, but I don't recall that it strongly underpredicted the shaking.

Maybe you're thinking of Somerville's argument that directivity could make motions much worse than expected for places in line with the direction of rupture. That argument is probably right.

John