Looking at your cylinders around arbitrary minor axes, I think there's a key point you're missing here: foreshortening. In your submission, there are two kinds of mistakes:

If you look at cases like those on this page, you're ending up with a lot of cases where the far end of your cylinders (the end not filled with hatching) is both wider in degree (which is correct) and larger in overall scale (which is incorrect, and breaks the key rule of perspective, that as we move an object farther away, it'll be perceived as being smaller). This mistake comes up a lot in your work, and I saw a lot of it both towards the beginning of this set, and further in. Strangely enough, there were some pages in between that were far better - like these (aside from 23 and 19, these are demonstrating proper foreshortening).

The other issue comes up on pages like this one. It's technically the same problem, but instead of having the far end get larger in scale, you're keeping the scale shift consistent from one end to the other, and only shifting the degree. This suggests that the side edges of the cylinders have a vanishing point at "infinity" (allowing those side edges to remain parallel on the page).

As discussed back in Lesson 1, the vanishing point only goes to "infinity", resulting in lines that are parallel in 2D space (as they're drawn on the page), when the lines themselves in 3D space are running perpendicular to the viewer's angle of sight. To phrase that a little more simply - your cylinder's side edges are only going to be parallel on the page like this when the cylinder itself is running straight across your field of view, from left to right, or top to bottom, as though the cylinder itself were sitting on the surface of the page itself. As soon as you rotate it even slightly into the depth of the scene, the side edges will start to converge towards a concrete vanishing point.

Because this exercise is to rotate the forms randomly and freely in space, that tiny bit of rotation towards or away from the viewer is basically guaranteed - so you have to assume that there is going to be some convergence to those edges that are parallel in 3D space.

Foreshortening itself manifests in two kinds of shifts, both of which we've already referenced here. One is the shift in scale, where as the far end of the form moves further from the end that is closer to the viewer, it's going to get smaller overall. The smaller it is, the more physical space there is between the two ends of this 3D form. On the page, we can only literally see the distance that exists from left to right, top to bottom, etc. The distance in the depth dimension of space, towards and away from the viewer, is "unseen" - therefore we rely on foreshortening to inform us of just how much more distance is being hidden.

Foreshortening also manifests in the shift in degree, where the wider the far end is (proportionally speaking - something can be smaller in scale but wider in proportion) the more of that unseen space is conveyed to be present. This also means that you cannot have a more significant shift in scale, but a minimal shift in degree, or vice-versa. They have to remain consistent, to convey the same information to the viewer, and to reinforce their assertions rather than contradict them.

In the most recent ellipses video for lesson 1, I explain with props how ellipses change as they move around in space, and how the relationships between the close and far end of a cylinder behave, so be sure to check that out.

Moving onto your cylinders in boxes, these are certainly better - there are definitely more notable convergences in the edges of your boxes, and as you progress through the set they continue to get tighter and improve. One thing that does stand out to me however is that I'm not sure you're actually checking the "true" minor axis of your ellipses after the fact. So for example, here are the minor axes of one of your cylinders. Checking this is important, because each ellipse has 3 lines that we extend (the minor axis and the two contact point lines), checking how far off they are from converging consistently towards the box's own vanishing points. The closer they are, the closer the ellipse is to representing a circle in 3D space, and therefore the closer the box's face is to representing a square in 3D space.

That is ultimately what this exercise focuses on - developing a more intuitive sense of how to construct boxes that feature two opposite faces which are proportionally square, just as the box challenge helped us develop a more intuitive sense of how to draw boxes whose edges fall into 3 sets of parallel lines. Now, despite not checking your minor axes, there has definitely been improvement in the consistency of those line extensions, which is good to see. Being sure to check your minor axes as well will help improve that further.

So, as it stands, your cylinders in boxes are alright, but the first set have some serious issues to address. I'm going to assign some revisions below.