In that case, I think it's up to the person in question to indicate that she needs someone else to stand in for her as she can't make a valid judgement. Anything else and you're overruling someone on potentially shaky grounds, which especially in the case of traumatic experiences is just something you shouldn't do.
Essentially, I disagree. But I think the most important thing to keep in mind, really, is that
framing matters. Humans are only, ever, imperfect recorders, and the way you frame the question can lead the answer. kaijyuu's point about "Have you ever been sexually harassed?" being a poor question is a good one. It has fuzzy boundaries, leaving too much up to interpretation, vague memories, and imprecise definitions, and narrows down their experience to a simple yes or no question. It's making a lot of unsafe assumptions and pretty much guaranteeing an increase in instances of reliable testimony. If you're working with statistics, you want consistent understandings of the variable being measured, and that's not a good way to achieve it.
It's why psychological studies don't just ask "Have you ever had periods of depression?" but rather require you to fill out a many page sheet of specific instances and details you are more likely to remember accurately, and then deduce from that whether or not you've had a period of depression major enough to warrant looking into.
Even that is imprecise. But when you're dealing with something with boundaries as fuzzy as 'sexual harassment' in the popular consciousness, at the very least you should define what you mean by it before asking the question, and allow the person to formulate their response within that framework. That's not "overruling their judgements" or arguing "someone needs to stand in for her" - it's simply making sure that communication is as intended. That's at the very least.