Try the frequency role in the distribution platform. At least with integer weights JMP gives quantiles that appears to be adequately weighted, at least they seem to be intuitively right. Fraction weights works too, but I don't know how JMP uses them in the quantile calculations.