This is always going to come down to opinion and I'm not sure one is clearly better. It's just like the teaching argument. Some programs require TAing and will tell you just how important it is for your future. Other programs don't require TAing and those will tell you what a waste of time it is.
I'm very glad I did 6 months of rotations before my PhD because it allowed me to settle on what residency I wanted to do. Going into clinics I wasn't sure, but the experience soured me on most of my top interests. If I went into my PhD not knowing what I wanted to do clinically, I wouldn't feel nearly as confident about merging my research interests with my clinical intersts.
Many will say "Well the goal of getting a PhD is general training". I'm not saying I disagree, but I think doing research in the future clinical area gives you a huge leg up not just for residency, but also for beyond. Then again, there's not so much of the usual cell and molecular biology in my area of interest. If you're planning going into IM, peds, or pathology (the top 3 MD/PhD specialties), all of what I said may not be so important because you can pretty much work in any cell/molecular biology lab and get the training you want.
Just between you and me (and everyone else in this forum), I'm beginning to think the best time to get a PhD is residency or post-residency. That's the logical extension of my argument anyways, and it makes alot of sense to me. But while we're on the topic, why not do the full year of clinics before getting the PhD? Oh yeah, because the drop out rate from the program will be too high. I guess the 6 month thing is a good compromise maybe?