Why do we even talk about any of these percentages like they're even remotely acceptable?
If your keyboard only produced the correct character 80% of the time, or your touchscreen discarded 20% of touches, or your speaker only streamed 80% of a song, not a soul would be okay with that.
Siri on your iPhone or Mac is a funny novelty. The moment it becomes the only way you can interact with a device? Who on earth would find that acceptable? Maybe you'd be willing to buy such a device if it was priced like a whoopie cushion or silly putty (so, you know, ~$5), but for $300?
And this isn't even getting into the fact that this basic test is EASY! We're not expecting the devices to be productive contributors every conversation - these are tailored queries - softballs - meant to be easily within the realm of what the devices are supposed to be capable of doing!