Most of the comparisons (that I've seen so far) amongst Maple, Matlab, and Mathematica are either incomplete, inaccurate, or biased.
This collection of three articles [1 (1), 2 (2), 3 (3)] published in 2005 by Computing and Science and Engineering was much better than most, as far as accuracy and bias go. It does, however, describe using those products more for educational purposes (eg. the teaching of math, science, or programming) than for academic research or industrial use.
What might be of interest are accurate, in-depth, and generally unbiased reviews or side-by-side comparisons of the three Ms, taken from the viewpoint of functionality needed for particular computational or research disciplines.