I'm a bit leery of the "marking guide" checklist approach you've suggested: it strikes me as unnecessarily inflexible. The commenting thing is a good example: it started off as a reasonable guideline ("-1 for too few comments, +1 for many comments, etc"), but it's certainly plausible that "too few" comments might be just right (if the code is elegantly written and short enough to grasp easily, for instance). So the "must comment" rule got modified to something a bit more vague: "effective use of comments". The same thing happened with modules: "use CGI.pm" went to "use applicable core modules appropriately".

Giving points based on a definite scheme is sometimes appropriate for assignments in a course, where you have to quantify how well students do in order to give them a final mark (and credits), but here I don't really think it's a good fit. Instead, you might use a checklist like this as a guideline for code reviews, but give a more subjective rating (as well as a small list of pros and cons).