This dichotomy isn't the actual choice, though. You'd hire the person with straight A/A-'s (read: nobody cares about perfect grades) from Stanford/Berkeley/MIT/Caltech who also worked directly with some well-known professors, publishing papers that advanced the field, albeit incrementally, either building experiments or developing theoretical results.

In other words, the person who has demonstrated solid abstract knowledge (breadth/versatility) and actual ability to do research (depth/hands-on).