The very same paper is littered with examples of cases in which stereotypes affect the judgement of people when your clause ("only applies to people that you are completely unconnected to and have never interacted with") doesn't apply. So I'm not sure if you're being intellectually honest, or cherry-picking the most extreme circumstances examined in the paper and generalizing them. Just one of a multitude of examples cited:
> a child from a high socioeconomic background who displayed a mixed performance on an ability test was rated as higher in ability than a child from a low socioeconomic background with the same performance (p. 293, citing two different studies)
So if you honestly believe that the stereotypes associated with gipsies cannot impact their lifetime opportunities in a harmful way, please do read in detail the Kunda & Thagard paper. And by all means believe the paper.
This contradicts the evidence that: "When individuating information is clear and relevant, its effects are “massive” (Kunda & Thagard, 1996, yes, that is a direct quote, p. 292), and stereotype effects tend to be weak or nonexistent."
After the interview, you have individuating information. In fact, you probably have individuating information beforehand, such as the resume, maybe references etc.
So your claim contradicts this study. As I haven't seen any evidence for your claim, my tendency is to believe the study.
But certainly something that probably warrants further study.