The algorithm for CLUSTALW works by calculating the similarity scores as the number of k-tuple matches between two sequences, accounting for a set penalty for gaps. The more similar the sequences, the higher the score, the more divergent, the lower the scores. While The MUSCLE algorithm proceeds in three stages: the draft progressive, improved progressive, and refinement stage. In this first stage, the algorithm produces a multiple alignment, emphasizing speed over accuracy. This step begins by computing the k-mer distance for every pair of input sequences to create a distance matrix. UPGMA clusters the distance matrix to produce a binary tree. From this tree a progressive alignment is constructed. The 2nd and 3rd stage is focused on obtaining a more optimal tree by calculating the Kimura distance for each pair of input sequences using the multiple sequence alignment obtained in Stage one, and creates a second distance matrix. UPGMA clusters this distance matrix to obtain a second binary tree. A progressive alignment is performed to obtain a multiple sequence alignment like in 1st Stage , but it is optimized by only computing alignments in subtrees whose branching orders have changed from the first binary tree, resulting in a more accurate alignment.
Further, in terms of accuracy both MUSCLE and CLUSTALW show greater accuracy in alignment of multiple sequences but MUSCLE is advantageous in handling more complex dataset and sequence length greater than 1000 bp.
It also depends what type of sequences you use. If they are protein coding, than both methods should obviously produce the same alignment. Especially with closely related congeneric species.
If you use ITS or other rDNA sequences, then the two methods might produce quite different results because of the more erratic (or less understood) way these sequences evolve.
Generally MUSCLE is much faster but less accurate than CLUSTALW.