I'm afraid Noels's post was not well-worded and might be slightly misleading as a result.
 
Improving processing efficiency could conceivably allow you to reduce your audio buffer size, which would in turn lower overall audio latency. But if you don't proactively change the buffer size, the efficiency gains will just result in lower CPU usage at a given buffer size while latency remains exactly the same. The one area you might see "latency" reduced automatically would be the time between hitting Play and having the transport start running, but this is not what most people mean when they talk about latency.
 
In my experience, most ASIO drivers only allow you to change buffer size by factors of two up and down, so it's unlikely you'd be able to take advantage of say a 10% improvement in efficiency to lower your buffer and still have the same CPU load in a given project. WDM drivers might allow small enough changes to take advantage, but with the inherent variability in CPU load over time, and the need to maintain "headroom" for spikes in the load, this isn't really likely to be feasible, either.
 
The main thing you get from processing efficiency improvements is the ability to run more synth and FX plugins before you have to start freezing things to get the project to run reliably at a given buffer setting.
 
EDIT: If you feel that audio latency has actually been reduced, you're likely just experiencing the Platinucebo Effect.