Hello,
The ASHA guideline is based off of the ANSI Standards, but written for the public- facing nature of the ASHA guidelines.
You are, on a large scale, correct that, from a psychometric perspective, the way that tones are presented to the user (the Houghston-Westlake process to obtain threshold) helps significantly to increase reliability. All automated systems use this method electronically, and so it is mostly up to the manufacturer whether the validity is maintained by alternating rate of presentation/gauging latency. The less intensive the microprocessor, the more likely it is ‘trackable’/predictable as people have mentioned.
This is usually acceptable risk in an Occupational setting because there will eventually be an assessment by an actual person at ‘some point’.
The reason that ‘even if you are unsure, still respond’ is stressed is because the volume/tone increases used should not be heard 100% of the time (floor effect), but 50%. If you are unsure, but still respond, it may be right at threshold or not-the end result should be the ‘proof’ even though there will always be exceptions.
Hope this increased uncertainty pulls back the curtain a bit!