My preference is usually to go with Balanced Overlap (B0). Complete Enumeration (CE) is great if respondents are robots, but when you have respondents who are humans you need to take into account their psychology: humans have lazy brains and those lazy brains look for ways to make difficult choice tasks easy. One way to make choices easy is to use a heuristic like "always choose the lowest price." If a respondent does this and you use a CE design, the only thing you can possibly learn about that respondent is that she likes low prices. With BO designs some choice sets will have overlap that will require respondents to step outside of their heuristics and as a result you will have more opportunity to learn more about their preferences.