I can well see how those factors would distort the test. Indeed, I'm not sure if "preference" is really the right word for what they're testing. What I found ominous was not just the result, but what I observed phenomenologically while taking the test.
When trying to associate "white and good" or "Jewish and good", I could feel myself coalesce them as a single concept, giving a single set of instructions to my fingers. When trying to associate the opposite pairs, I could FEEL myself fighting an uphill battle.
It wasn't that I was silently objecting to myself "but black people are bad" or "but other religions are bad", obviously. But it felt like an unnatural association, a strange and unfamiliar way to parse the world. This doesn't say much about "preference" as an individual, personal act of the will. Rather, it's a way of viscerally feeling the effect of culture. But that's almost scarier. In every instant that I look at a given human being's face, the insidious, silent, unremarked voice of culture is whispering in my ear, associating that person "naturally", almost "linguistically", with good or bad.
Not that this is news, but it's a frightening demonstration of it.