Big Jeff, this question comes up in various ways on a fairly regular basis.
Sets have inputs, of course, and output, which is the screen. In RPTVs, the arrangement of pixels for the screen is not identical to standard HDTV inputs, so images need to be mapped to the different screen layout so the images have the proper aspect ratio. The screen might be 1364 x 738, for example, on a good set. It's not exactly 720, so the image is redistributed across the pixels.
"Upconverting" means that the input signal is converted to a standard one that has more pixels before being mapped to the screen. If the signal is non-HDTV, such as 480p or 480i, some take it to 720, some to 540, before mapping it. They all can map 1080 down to their screen layout.
All that said, is it inherently better to upconvert to 540 or 720? Since you're starting with the same information from the input, it might not make much difference. Many of us assume that upconverting to 720 implies that the processing is more sophisticated or that the screen has a greater number of pixels. It certainly sounds better to talk about it. Is it really better to upconvert to 720, which the set has to be able to map from? Or is it better to have a mapping algorighm specifically for some intermediate level or even for 480 directly?
Either could be better.