It isn't a "moral" issue. It is about the usefulness of the benchmarks. If you're using them to make a recommendation about which programming language to choose, it is important to have a sense for how well the results generalize to the kind of programs you will be writing. Using unrealistic hacks targeted specifically to each benchmark is contra that goal.

If, on the other hand, you're just treating the benchmarks as a fun competition akin to code golf, then by all means, hacks ahoy!