Starting with your cylinders around arbitrary minor axes, your work here is by and large coming along well. I can see clear signs that you're consciously applying the ghosting method to each and every mark. There are still some very, very subtle signs of wavering at times, but it's very very slight and should continue to disappear with practice, as long as you continue to focus on executing your marks with confidence and doing so using your whole arm from the shoulder.

One minor point I wanted to mention is that when you're deciding how much of a "degree shift" to apply between the ellipse closer to the viewer and the end farther away, keep in mind that both the shift in the degree and the shift in scale (where the far end is smaller overall due to the convergence of those side edges) are manifestations of the same thing: foreshortening. Meaning, they will generally work together, in tandem. If you have a lot of convergence to those side edges that results in the far end being a lot smaller in its overall scale, then it should also increase its degree a fair bit to match. When we maintain the same degree while allowing for a larger scale shift, it'll make the result feel a little stiff and awkward, as we can see here (although don't worry if it doesn't look immediately obvious, it's one of those things that your eyes get better at detecting, although the viewer will often pick up on something being a little weird).

Carrying onto your cylinders in boxes, your work here is fairly well done, although I do have one suggestion that I'll share in a moment. All in all however, you're taking great care to apply your line extensions correctly (though I can see some extra ones where you appear to be connecting the ends of the contact point lines from the different ellipses which isn't necessary, and shouldn't be included just to avoid any unnecessary complication).

This exercise is really all about helping develop students' understanding of how to construct boxes which feature two opposite faces which are proportionally square, regardless of how the form is oriented in space. We do this not by memorizing every possible configuration, but rather by continuing to develop your subconscious understanding of space through repetition, and through analysis (by way of the line extensions).

Where the box challenge's line extensions helped to develop a stronger sense of how to achieve more consistent convergences in our lines, here we add three more lines for each ellipse: the minor axis, and the two contact point lines. In checking how far off these are from converging towards the box's own vanishing points, we can see how far off we were from having the ellipse represent a circle in 3D space, and in turn how far off we were from having the plane that encloses it from representing a square.

Now the main thing I want to suggest is that you try and push the rate of foreshortening for your boxes a little more to avoid falling into the trap of making your boxes' foreshortening so shallow that they effectively stop adhering to the rules of perspective and shift instead to being more "isometric" (these terms are explained here and in the video at the top of that page). You don't push quite so far as to end up making that mistake (which only happens if we try to force all our vanishing points to infinity), but you do get close enough to it that it could result in some confusion. Just be sure to push the convergences a little more - this will also serve to make the line extensions more useful, as you'll be able to judge your line extensions more easily.

Anyway, keep those points in mind, and I'll go ahead and mark this challenge as complete.