Yeah, I'm wondering, too. MacWorld did exactly that:
No they did
not do exactly that.
The miniDisplayPort (mDP) display is plugged into the external TB drive;
not the TB monitor.
I suspect it makes a significant difference. What they are probably trying to stop if people using the TB Display in legacy contexts ( to a display only computer). If the Promise Pegasus drive picks up the DisplayPort (DP) data traffic and routes it to the TB Display, then it will get routed along to the host computer. It is clear that it is not meant for the TB Display and the TB Display does not have to mess locally with DP legacy mode.
There are two huge user failure modes potentially with the TB Display.
1. The users treats it like a normal DP display and plug it into something not TB. The ports don't work but they call , complain, moan , whine about the ports not working.
Making it so that system doesn't work in that context solves that problem 100%. It won't turn on so at least some users should catch a clue they are doing something wrong.
2. Users tries to treat two DP connectors like they would on any other mainstream monitors. That means you can hook the monitor to
TWO different computers. Normally that what that means. Millons of people have done this. Well that is not allowed for a TB network. It is like USB. There can only be
ONE computer. Perhaps several smart peripherals but just one host computer with the highest level PCI-e switch and the GPU.
So the likely failure in the document is that it does not make clear the "work around" for the problem. You must buy another, non-display, TB device in order to hook up you DP monitor. It tries to do that by stating you should hook up the TB Display first and then add TB devices
after the TB display. But it doesn't state what that "buys you".