Didier Stevens “benchmarked” the efficiency of ALSR as implemented by the EMET tool.

The conclusion is that it is pretty weak, whereas I thought it was on pair with true ALSR (as advertised). Very instructive.