That's a fair observation which proves the well-knows fact that microbenchmarks are triky to get right. Thanks for sharing! I'll put a link to your comment into the post to make it more precise.