[thaumasiotes]

Yes, the structure of the test is designed to produce a fake result based on a spurious premise.

It wants to do a binary search in the color space to find "The Boundary" between your personal concept of green and your personal concept of blue.

No such boundary exists. If you abandon the idea of binary search and just present a bunch of similar colors in no particular order, you'll get more realistic results, which look like "you called color x 'blue' x% of the time, and 'green' (1-x)% of the time". You could even display a grayscale histogram of the blue-green continuum according to the odds of any particular point on it being labeled "blue" or "green". That would be fun.