bulgie 

Alarming that ShelBroCo would show a picture of doing it wrong. (I am admittedly too picky by half, but why not do it right? It's not any harder.)

The string should come off the dropout at a point that matters, namely where the wheel will sit, the inner face of the "land" for the axle. This shows the string coming off a part of the dropout that's not involved with holding the axle, and it's different between the right and left (that knot can throw off the string's centerline).

Even if you do it right, you might want to measure and compensate for the curve in the seat tube from brazing on the front derailer attachment. All tubes have some runout of course, nothing is perfectly straight, but the F. der. BO puts a pretty decent curve in the tube, more than any runout I have measured in an unbrazed tube.

So the string method is not the ultimate alignment, but it's decent as a shade-tree mechanic technique for those without more rigorous methods, or even for a pro framebuilder in too much of a hurry for a complete alignment, say for a frame that isn't valuable enough to warrant the complete tear-down and rebuild. Just run that string in the right place, and try to see how bent your seattube is!

Note this string method (done right) is equivalent to the Park ***-1 tool shown in the OP; they measure the same thing. So the ***-1 (or -2) is also, in my opinion, not a "be-all and end-all" pro alignment tool, but it is very fast and easy to use. I have one and use it, but I keep its limitations in mind.

EDIT: the asterisks were put there by the censor. They replace the name of the Park Frame Alignment Gauge.
I, for one, welcome our new robot overlords!

Mark B