I think I see what you mean. I believe the way you approached both is correct, although the latter approach is more in line with the instructions (the enveloping lines) rather than the points. When using the points, you technically treated other subdivision lines that were in the same area/connected to the dots in the way you used the enveloping lines, but if we were just given a series of dots to connect, if we didn't first actually connect them with straight edges, we'd be left estimating a lot in terms of the actual behaviour of our final lines. Having straight edges to work from allow us to focus more on how the structure is being rounded out - kind of like carving something from a block of wood, where it starts out very planar and straight, and is only smoothed out in key areas towards the end.