Hopefully this brings some resolution... both to my initial question of what a realistic expectation of flatness is (spoiler alert: flatter that what I got!) and to how significant flatness is.
From carefully reading about the adventures of the waterstone crowd, the magnitude of out-of-flat is far less important than the shape of that out of flat. And what you want to sharpen. Visualize sharpening a plane blade, maybe even with one of those jigs that you roll across the stone. If the out-of-flat is shaped like a skateboard half pipe that isn't so bad. The angle of the blade doesn't change much, not a lot of concern. HOWEVER, if the out-of-flat is bowl shaped, there's a problem. The wide and straight plane blade will assume the shape of the stone - a slight U shape on the blade. That isn't good. Ditto for chisels. Worse, that "U" in the stone is not consistent - on either end of the stone it's barely there and it gets big toward the center. But if you're sharpening a curved knife edge, a bowl shape may not be a big deal at all. There are apparently very competent waterstone people who seem to think of stone flatness in terms of 16ths of an inch for knife sharpening. And you can probably think through the shape of a stone out-of-flatness and find all sorts off scenarios where it does and doesn't matter. < all my non-expert observations >
As for what a realistic expectation of flat is... I contacted DMT via their web page a couple days ago. No answer (yet?). I called the establishment I bought it from (sharpeningsupplies.com) and spoke to a technician. I explained how I measured the flatness and what I'm seeing. It took him about 0.5 seconds to say that was definitely beyond what it should be and he's sending me a replacement w/ a return shipping label for this one. I asked him about breaking it in first and he said I'm welcome to try and that will knock down a lot of the roughness you feel on a new stone, but it's not likely to change that amount of out-of-flat. We did talk about DMT's coyness about stating a flatness tolerance. In the tech's experience these stones should be in that ~1 thousandths territory, but nowhere near 4.5 thousandths. They (DMT) do check for this, but sometimes one slips through. They're happy to replace it and disappointed it happened. < this is why I avoid Ebay! >
Bottom line, out-of-flat may not matter depending on the shape of out-of-flat and what you're trying to sharpen. But if you lay down ~$140 on a diamond plate it'd better be flatter than 4.5 thousandths.