I don't think that any of you answered the question. I'll try as best I can, but it'll take a while.
I'll use my pocket digital camera as an example that'll lead up to the answer (follow along, there is a reason). With it, I can choose to take pictures at the highest resolution (2048 x 1536 I think). When I zoom in to 100%, meaning I view the image exactly how it was recorded to the CF card pixel for pixel, the edges aren't so sharp. Perhaps it's because of anti-aliasing, but for whatever reason, the edges aren't super-sharp. It's sharp, but not unbelievably sharp. The sensor that's on the camera's known as a "Bayer" sensor. Bayer-type video capuring devices, the type used by virtually every single video camera out there, aren't terribly sharp per-pixel at the "pixel-peeping" level.
Now let's say I downres the in-camera setting to 1600 x 1200 like I have on occasion. You would think that when I zoomed in to 100%, I would have super-sharp pixels because of the downres. But no -- the edges are similarly "blurry" as if I had taken them at full resolution. I don't know why -- but again that doesn't mean it's not so.
The same effect's probably going on with these cameras. The setting's put at 720p and the edges are as "blurry" pixel-for-pixel as the camera would capture it in 1080i. BUT, if that 1080i picture is externally downrezzed by an external method, then this bypasses the camera/sensor's processing unit, and allows a "per pixel" sharpness that is much sharper than it was natively captured -- and sharper per-pixel than the native 720p out of an ESPN-HD cam.
I hope that made sense. Sorry if it didn't.