I have my doubts about it.
Each system, each architecture, each configuration is different when it comes how well performance is optimized and exactly what the bottleneck is (by definiton, there has to be a bottleneck).
There are just way too many variables to say such a thing.
There is also the problem of looking at a system, ie a hardware/software combination, and treating it as if the software didn't exist or has no effect on performance.
More memory can increase the time certain tasks take, there's no doubt about it. Startup on a Mac, for example, involves a memory test. It's obvious that startup must take some small amount of time longer if more memory is installed. But using the computer is much faster with something above 100% of the optimal memory for a given task.
I don't see the point of his comment.