These voice-activated chips can be trained (as seen in a lot of other phones), but I'm not sure the software-powered Siri can be trained.
So, I think the error rate is simply not low enough to make conclusive claims about what it might or might not do.
I know Android users might not understand this limitation, but there it is.