Optimal value of decision variable leads to inconsistency

Question

$\epsilon$ is a random variable with support in $(0.8,0.95)$ and pdf $f(\epsilon)$. The following equation arises out of a business problem: $ENP=800*A*E(\epsilon)+ 9000 - 250*\int_{0.8}^{\frac{0.75}{A}}(0.75-A\epsilon)f(\epsilon)d\epsilon- 600*\int_{\frac{0.75}{A}}^{0.95}(A\epsilon-0.75)f(\epsilon)d\epsilon$ On substituting the distribution of $\epsilon$ and simplifying we get, $ENP=700*A+ 9000 - \frac{250}{0.3}*(\frac{0.75^2}{A}+A*0.8^{2}-1.2)-\frac{600}{0.3}*(\frac{0.75^2}{A}+A*0.95^{2}-1.425)$. Here $A$ is a decision variable. On optimizing above equation w.r.t. $A$ we get optimal value $A^{*}=0.9863$ and $\frac{0.75}{A^{*}}=0.76$.Why is this happening?$\frac{0.75}{A^{*}}$ should lie in the range of integration i.e. $(0.8,0.95)$.