You can think of the exercise as being broken down into two parts- firstly, there's what we focused on with the box challenge, which is simply drawing arbitrary boxes whose lines converge fairly consistently towards the same vanishing point. It doesn't actually matter if the foreshortening itself is dramatic or shallow - either way, it comes down to thinking of all 4 edges simultaneously, and considering how they need to be angled to converge together. The best way to address this is just to do more freely rotated boxes (this should already be one of the many exercises you cycle through in your warmups, so you can keep sharpening those skills).

The second part is the addition of the ellipses, which effectively tests the proportions. Based on the ellipses' own line extensions, and their relationships with the box's line extensions, you'll be able to make adjustments to which proportions you choose to use, based on the general orientation of your box in space.

Of course, if your box's own lines aren't converging entirely consistently, then that will mean you'll have to eyeball the relationship between those line extensions and the ellipse's with some margin for error. You can still of course progress through this exercise meaningfully without perfect boxes - in fact, it's not really expected that your boxes will be perfect - but it means that you have two distinct things to deal with, and remembering that they are separate issues will help you address them each in turn.

As for the why behind those aspects of foreshortening, the best explanation I have right now is in the more recently updated Ellipses video for lesson 1. Here I actually use a little prop of two cardboard discs connected by a toothpick to show how the degree of the end closer to the viewer and the degree of the end farther away will shift in degree.

The rate of foreshortening itself is largely determined by the form's proximity to the viewer and the form's own length, so in this video you don't get a lot of variation in that. You do however get to see the third factor, which is the orientation of the form relative to the viewer, with that foreshortening getting more notable as we turn the cylinder to face the camera. Hopefully that demonstration will help.