The reason a long menu shortens unevenly is that entries and owners are not the same thing. A COM context menu handler is a single registration, but the code behind it decides at display time how many items to draw. An archiver commonly shows a top-level entry plus a submenu of several operations. A cloud drive might add a share action, a link action, and a version-history action. All of that is one handler with one identifier.
This has two practical consequences. First, effort per visible entry removed is wildly different depending on the surface — disabling one handler can be worth five rows, while a classic verb is worth exactly one. Second, it explains a common and confusing failure: you remove several entries individually, they disappear, and then after a reboot or an app update they are all back together. They were never independent. Whatever you did addressed the display rather than the registration, and the registration reasserted itself.
ContextCleaner surfaces the identifier behind each entry, so shared ownership is visible rather than guessed at. When you disable an add-on, every entry it contributes goes at once, and the change is recorded as a block-list entry keyed by that identifier rather than as a deletion of the app's own registration. The app remains fully installed and functional; it simply no longer draws into the shell menu. Re-enabling removes the block-list entry and the entries return.
The New submenu deserves its own pass for a different reason: it is a category rather than a scattering. Every entry in it is the same kind of thing — a document type offered for creation — so a single focused review handles all of them, and the decisions are easy because you either create that file type or you do not. Handling it separately from the main list means you are not making that judgement repeatedly while also thinking about unrelated entries.
What is left after those two passes is usually a short tail: individual classic verbs from specific tools, a couple of modern-menu items from packaged apps. Those are worth doing one at a time, and at that length it is quick. The order matters because it front-loads the passes with the highest ratio of entries removed to decisions made.