25% is a funny number.
25% is the number of people who simply would not go along with the Asch Conformity experiment.
25% is close to the number of people who would not go along with the Milgram Authority experiment. We might speculate that the extra stress of the pain and suffering made more people than usual relent (35%).
25% is the proportion of expressed recessive traits in a two-allele gene. 
The funny thing about recessive genes is that they cannot be simply bred out.
In a typical Punnet Square, the quarter of the population expressing the recessive trait can be prevented from breeding, and half the population still carries the recessive allele, so they can still interbreed and bring out the recessive traits. If you do the inferential statistics and assume a random distribution, the recessive trait will often still be in a majority of he population.
Recessive traits cannot be bred out.
Perhaps more to the point: killing all the non-conformists hasn’t ever actually made the population conform. Perhaps it’s because non-conformity is recessive.
