This is similar to a common technique applicable to potentiometers used as variable resistors, and I'm guessing that it is done for similar reasons. However, it probably isn't a good idea to implement as drawn.
Consider a potentiometer connected like so:
simulate this circuit – Schematic created using CircuitLab
Ideally, the connection between the wiper and the left end of the potentiometer would do nothing, since the portion of the potentiometer to the left of the wiper is shorted out at all times, and without that connection would be open circuit at all times.
In practice, a potentiometer may lose connection between the wiper and track momentarily as it is turned, or an extremely worn one have “dead” spots of no connection even when not being turned. Therefore, the extra connection ensures that the resistance between A and B is always no greater than the end-to-end resistance of the potentiometer. This is important if a too-high resistance or open circuit could damage the circuit (for example, if A is ground and B is connected to a voltage source would, by itself, raise point B to a voltage that would damage other components), and more generally means that the potentiometer's "output" cannot become "out of range" in the event of dead spots (unless the potentiometer's track/winding breaks entirely).
In your case, you do not have a potentiometer but a switch — therefore, unless it is a make-before-break type (which might be a good choice anyway), it will always behave like the worn potentiometer, causing J1 to become open-circuit as you make the change. Thus, the extra connection ensures that the antenna is always connected to J1, regardless of what is done with the switch. It might present a poorly-matched load, but not a pure open circuit.
But, there's a catch — instead of shorting out portions of a resistor we're shorting out portions of an inductor. All turns of an inductor are coupled together (though more separated ones less so), so there are now shorted turns, which will waste power in eddy currents rather than radiating.
Therefore, the presented design with a tapped inductor is not a good idea. You could fix it by using separate inductors for each stage, arranged to minimize coupling between them (perpendicular axes, keeping them apart).
To address your broader question of what is the proper way to electrically wire in a loading coil to a vertical antenna — I don't have any personal recommendations, but I do think that the extra wire that you were puzzled by is plausibly a good idea. Whether it's worth the complications of implementation, I can't say.