Merging an HDR from a single DNG graded two different ways is totally pointless and won't give you anything that the original DNG didn't contain. You have to have DNG's that were recorded with different Exposure Values to get anything useful. The main concept behind merging HDR's is that a single image cannot capture the full luminance range, clipped white values for example. So if you have an under exposed DNG to capture the shadows and and over exposed DNG to capture the highlights merging them together will give you an HDR that has a full range of values from light to dark.
And that's the beauty of RAW, it contains information that's not within the Dynamic Range of what the picture is showing.
Sure, you can throw one DNG into Photomatix and see where that will get you but it's the way photomatix interpreted it that makes it a different story.
Also, you could try to make one overexposed DNG with some noise reduction applied to it and a normal under exposed picture, this will get you better results than just throw one DNG in there.
And yes, it's better to capture multiple exposures, but with RAW video we simply don't have the power for it, so this double processing is a very good alternative!
Edit: This guy explains it much better:
http://captainkimo.com/single-exposure-hdr/