You are referring to Schelling's model, and the threshold he found for segregation wasn't 1/3 but 2/3 'wrong', i.e. that everyone prefer not to be in a minority of less than 1/3. This was just was striking about it – that even if every individual is happy with being in a minority, the population can still segregate. I'm not aware that he ever referred to the degree of preference in his model as 'mild', or small. It's relatively small, compared to what you might otherwise expect – I think most people would have guessed that it took a preference of 1/2 to produce self-segregation.