Our take is more that if you can recognize what is wrong with the code on multiple levels then you can most likely write good code.
This is easier to check on non-trivial code when done through a code review then by getting the interviewee to write the same kind of code that every other interview looks for. The submitted code sample does this as well.
I'd also want people on my team to be able to successfully refactor legacy code just by looking at it and being able to identify where things can be better cleaned up and tested.
I am completely behind code review and refactoring. I am our company's Chief Refactorer (a title I just made up, but it's accurate), and after instituting code review a few months ago, it has made our development process so much better. At this point, I hate merging in a feature that I've written without having someone else review it.
So that's not my issue. I was pointing out that your interview process seems to involve only code review, based on your comment above. Put yourself in an interviewer's shoes: in two different instances, you've been presented with bad code and asked to code review it. Personally, I would be concerned that (1) your company writes bad code, and (2) I'm going to be expected to spend all my time reviewing and fixing other people's bad code instead of writing code myself.
Now I doubt that's actually the reality of working at your company (at least I hope so!), but first impressions are very important, and that's what your interview process would tell me if I were interviewing there.
Our take is more that if you can recognize what is wrong with the code on multiple levels then you can most likely write good code.
This is easier to check on non-trivial code when done through a code review then by getting the interviewee to write the same kind of code that every other interview looks for. The submitted code sample does this as well.
I'd also want people on my team to be able to successfully refactor legacy code just by looking at it and being able to identify where things can be better cleaned up and tested.