why pdf fit changes if 1 data point is dropped

I am trying to fit p(x)=a^2*x/(1-bx)^3 * exp((-a^2*(x/(1-bx))^2) 0<x<1/b to my data vector x containing 300 points.

I get a reasonable fit (visually) and estimates of a and b.

Now if I drop min(x) and try to fit the same pdf to the 299 points, my fit changes rather significantly, so do the estimates of a and b.

Is this acceptable? Can anybody explain reasons behind this?

[Walter Roberson]

How different is min(x) from your other points?

When you write that 0 < x < 1/x then is that indicating 0 < x < 1 ?

[Russ Adheaux]

Sorry Walt, I had a typo in the question. My limits of x are 0<x<1/b (I corrected it in my original post. I had 0<x<1/x incorrectly)

Also my min(x) is not an outlier. Its in the range of next higher values of x.