Average Accuracy

//We are trying to find a mode estimation for a continuous
//distributed set.
//Anyway I am still working on a solution called 'half range method' and I came across a c# peculiarity.
//I have a recursive binary search (essentially) and we are trying to
//performance optimize it by instead of recreating the array of floats each pass,
//use indices to narrow down the search.
//Riddle me this:

float test = input.Average();

int count = (top - bottom) + 1;//number of elements in this iteration
int pos = bottom;
float average = 0f;//working average
while (pos <= top)
{
     average += input[pos];
     pos++;
}
average = average / count;

//my average method returns 2 places less accuracy than c# .Average()
//example:
//0.0371166766 - c#
//0.03711666 - my loop

//125090.148 - c#
//125090.281 - my loop

//Any light you could shed on the subject would be much appreciated!
//No rush.
//Thanks in advance
// - ST