I worry a bit about this kind of surveys. By this standard, I've faced sexual assault because a drunken guy started petting me once (I'm a straight guy). By any reasonable standard it was a funny incident, but in a strict interpretation of a "have you ever" survey, I'd be a sexual assault survivor.

Could be the same here. Say that in 1/100 of interactions between a upper and lower caste there is some form of discrimination, regardless the intensity. I'm too tired to do the math, but it seems reasonable that would lead to a high percentage of people ending up answering yes to a "have you ever been discriminated" question.