Run of the mill laptops shouldn't be factory undervolted at all (too much work to ensure stability) and as far as I know, in the case of Dell, not even the Alienware line of laptops, where you might expect it, is factory undervolted. Undervolting potential is always a lottery unless the manufacturer specifically bins the chips. And even then, volume can be a problem so the top bins might have to be pretty wide. They might be avoiding the worst for Alienware, but I don't think they have a top bin for them (and certainly not for the XPS). Frankly, if you're interested, forums are the better source. It's a question of statistics. You need samples. A lot of them.
The clearly beaten statement is in a paragraph talking about continuous Cinebench performance. It's hardly surprising it pertains to that particular test and not the first run. And while I value these longer tests, I wouldn't say it's relevant to the typical user. It's not like it's a gaming laptop where I would want to ensure the benchmarks are long enough as gaming sessions are easily tens of minutes if not hours long. A five minute benchmark might be misleading. For ordinary use, how many people actually run tasks that take 10, 15, 20 minutes? On a tiny laptop that's trying to cook itself. Usually, you get bursts like loading up a web page. Video playback is the typical continuous load for which you have hardware acceleration (without it, the laptop would be dead in no time).