quoting nevent1q…qrprI don't see a point to the mutual-following-based web of trust except maybe sheer simplicity. A better web of trust wouldn't purely focus on who trusts who.
It would factor in how much you trust people and what topics you trust them on by letting you shift the priority list of what npubs you trust.
It would also factor in which content people do and don't trust from each other in general by using content-based judgments instead of only npub-based judgments at the furthest degree of separation. In other words, putting an npub at your highest priority slot would not filter out what npubs they disapprove of and show stuff from what npubs they approve of. It would show you what content they approve of and filter out what content they disapprove of.
The "approving and disapproving" has different meaning depending on context, since there are different uses for web of trust. In the context of feed curation, posts they approve of could mean posts from anyone they follow and posts they don't approve of could mean posts they've reported. That's kinda like the existing thing with mutual follows. But in the context of a reddit clone, the web of trust could be like a weighted karma score based on upvotes and down votes. In the context of a wiki, upvotes and downvotes would make sense as accuracy judgements there too.
It isn't about mutual follows or prioritizing the npub itself in a crude way. The trust graph is context-dependent (as you rightly point out is the best way to go), subjective, and follow-agnostic: https://github.com/nostr-protocol/nips/pull/1208