OK, I suggest doing a quick plot:
You can see that even though it appears that x1 and x2 are different, they are actually the same. This is logical, since you are obtaining the FFT of the same signal. The only difference is that since in x1 you use 128 points and in x2 you use 64 points, the FFT of x1 has a better resolution in the frequency domain (look here for a detailed explanation).
The division between 128 and 64 is just to normalize them and make a fair comparison (try plotting without the division and you will see what I mean).