Simply put, the assumption that both x and y are equal to the same value is incorrect. One or both have accumulated rounding error in some undescribed process.
% x and y are assigned the same value
x = 1.6;
y = 1.6;
x-y
% y has accumulated some rounding error
k = pi;
y = k*(y/k);
x-y
% the difference is nonzero, but they look the same
[x y]
% but that's just because the error is small relative to the nominal value
[x y]-1.6
Not all numbers can be represented exactly in base 2, just as not all numbers (e.g. 1/3) can be represented exactly in base 10.
