When you make a substitution to a new variable of integration, you can either a) fix your limits of integration by putting them through the function or b) leave them as "x=whatever" and then when you are done integrating, put back your original variable and evaluate. But not both, which is what you seem to have done here.
Let me see if I can make that clearer. Consider [itex]\intop_1^2 (x^2-2x+1) dx[/itex].
Now you can just evaluate this:
[tex]\frac{x^3}{3}-x^2+x \vert^1_2 = \frac{1}{3}[/tex].
But let's suppose you decide you want to do a substitution:
[itex]u=x-1, du=dx[/itex]
[tex]\intop_1^2 (x^2-2x+1) dx = \intop^?_? u^2 du[/tex].
Now you have two choices, you can move the limits from x to u by plugging into your substitution:
[tex]\intop^1_0 u^2 du = \frac{u^3}{3} \vert^1_0 = \frac{1}{3}[/tex]
or you can write:
[tex]\intop^{x=2}_{x=1} u^2 du = \frac{u^3}{3} \vert^{x=2}_{x=1} = \left[\frac{x^3}{3}-x^2+x-1 \vert^2_1 \right]= \frac{1}{3}[/tex]
Sometimes it's more convenient to move your limits of integration to the new variable and sometimes it's not. But what you did was convert your limits of integration to the new variable u. Then when you were done integrating, you converted the expression to be evaluated back to x, but plugged in the "u" bounds.