Coming from a real-time environment I recall asking that question back in the 1980s, and while I agree with the complexity argument, there could have been at least some baseline measurements for the simpler addressing mode cases.
I think we can be certain that DEC did measure performance, and that if the 780 had been significantly faster than its competitors then I seriously doubt that the marketing people wouldn't have advertised that.
IIRC the early versions of Unix for the VAX first used the CALL instruction to invoke functions, but then dumped it in favor of JSB because of the performance overhead that CALL entailed.
They certainly wouldn't have wanted to publish performance data for the 730. I had one and it was no match for an 11/73.