Some caveats: sometimes it will still want to stop at a green light, in my experience, and requires a manual override; and, if you're the foremost car in your lane at a traffic light, it won't begin moving on its own. I assume the same is true for stop signs.
I guess that video is intended to be a preview of what the current software could do with all the driver interaction safety switches off (no required hands on wheel, no requirement to confirm safety through intersections/turns, etc) and all the internal feature flags turned on (particularly: enabling turns and enabling Navigate on Autopilot on non-freeways).
Perhaps the driving engine was manually pre-trained for this particular stretch of road. Maybe what we see in the video is one successful run out of 100s that have been tried. We do not know the details. In my opinion it is reasonable to assume that it was a publicity stunt. Maybe if you try the same thing on a neighboring road two hours later, it would fail 95% of the time.
So far, the customers cannot use this level of autonomy. We can only speculate why but I believe that if it was usable in real world scenarios, Tesla would have enabled it. Since Tesla does not enable it and instead focuses on low hanging fruits like stopping on red lights, it appears that the technology is not ready yet. I think that is the main issue with Tesla's advertising about it. Not the name they use for it.
I can imagine that $8000 option of "full self driving" sells better than $8000 option of "restricted driving assistance limited to ideal weather and traffic conditions hopefully available in a few years", for instance. But at the same time I believe that using misleading product names or descriptions in order to increase the product's perceived value should not be tolerated.