You've made the classic so-called fencepost error. Think of the trapezoids as the gaps between posts of a fence, and think of the evaluation points as the fenceposts. There are n of gaps, hence n+1 posts. In your CompTrap, you have n+2 evaluation points: a, b, and n done in the loop. The loop needs to stop at n-1.
Also, in your test comparison integral, you are integrating from 8 to 1. You should go from 1 to 8.
You can compare your procedure with Maple's own trapezoid rule procedure like this: The call
CompTrap(f, a, b, n);
should produce the same result (*footnote) as
Student:-Calculus1:-ApproximateInt(f(x), x= a..b, method= trapezoid, partition= n);
This is a good way to test whether your procedure is correctly implementing the rule.
It is also good to compare against the integral done by Maple's more sophisticated methods, as you did. This is a way of testing whether the rule itself is good, not testing your implementation of it. You can also study the rate of convergence of the rule.
A matter of style in your procedure: Be careful of how you mix "pure" exact symbolic computation with floating-point numeric computation. Your procedure has 0.5 near the end. You should probably change that to exact (1/2).
Regarding your Simpson's Rule procedure: In addition to the fencepost error, there five other errors in that, three algorithmic and two syntactic. Two of the algorithmic errors are super obvious and the third is subtle; both syntactic errors are essentially the same. If you find any two of the five, I'll point out another two. Deal? Also, why did you change the parameter n from the trapezoid procedure to j in the Simpson's procedure? Maple doesn't care which you use, but the change may have clouded your thinking and led to one of the errors.
You can test/compare your procedure with Maple's own implementation of Simpson's rule by using the same Student command as above, but changing trapezoid to simpson.
(*footnote): After you apply evalf, there may be a small deviation due to rounding error. This may affect the last digit.