expdf fits a pdf to a distribution of points... it doesn't define the pdf at given x-coordinates. So in your code, you're fitting an exponential distribution to uniform data (the x-coordinates), which is of course not what you want.
If you just want a curve following the approximate shape of a negative exponential, then I recommend defining the curve explicitly, rather than messing with fitting:
x1 = -25;
y1 = 0.015;
x2 = 0;
y2 = 0.0025;
dy = y1 - y2;
dx = x2 - x1;
k = 1;
f = @(x) dy.*exp(-k*(x + dx)) + y2;
x = linspace(x1, x2, 50);
y = f(x);
You can adjust the k parameter to get a steeper or shallower drop-off in the curve.