Why does my engraving look patchy on polished aluminum?

The reflectivity is part of it, but I'd bet the curve is doing more damage than the polish. On a curved cover the distance from the lens to the surface changes as the head moves across, so anything sitting outside the focus range comes out faint no matter what the material is doing. That lines up with your symptoms: crisp where the surface happens to land near focus, washed out where it falls away. Rotating the piece won't fix that, it just moves the weak spot somewhere else. What worked for me on curved stuff was shimming the part so the worst area sat dead at focus, then running the job in sections with the focus reset for each region. Slower speed on the steep patches too, since the beam is hitting at an angle there and the energy spreads out.

On the scotch-brite question, yes, a light scuff absolutely gives the beam something to bite on and evens out the marks a lot. But you're right that it wrecks the mirror look, so try it on a hidden spot or a junk cover of the same material before you commit. If you want to keep the polish intact, the usual route is a thin marking compound for bare metal. You spray or wipe it on, run the job, and the laser bonds the coating to the surface instead of fighting the reflection. Wipes off with water afterward and the finish underneath stays shiny.

One more thing about the dithering. Polished metal makes dithered shading unpredictable because each tiny dot is basically rolling the dice with the reflection, so some dots just vanish and the gradient breaks up. Try fewer, bigger dots at a lower resolution and see if the shading holds together better. And I'd do one slower pass over two faster ones. A second pass over a half-marked area lands differently than the first, which is probably part of why your results are so patchy.

Still pretty new here but I did something similar on a curved flask lid last month, mostly flat except one domed edge and that edge is exactly where my engraving fell apart. The section-by-section focus idea makes sense to me but I'm not clear on one thing, how do you blend the sections so you don't get a visible seam where one pass stops and the next starts? Maybe that doesn't matter on engine covers if the design breaks it up anyway. I might test on my spare lid first before touching anything I care about.

Ran into the same headache on a domed tank badge. What helped was running a low power test line down the curve first and marking where the mark faded out. That showed me exactly where to split the job into strips, and I overlapped each strip a little so there was no visible line where one run ended and the next picked up. Slower head speed on the steep part of the curve too. If you do try a scuff test, rubbing compound on a rag brought my test piece back to near polish afterward, so a junk cover might not be a total loss. definitly do the test piece before touching the good one though.