docs: clarify architecture evidence boundaries

README.md

@@ -15,6 +15,11 @@ location so consumers can distinguish it from imported symbolic evidence.
 
     cargo run -- build ../fold --json fold.brick.json --html fold.brick.html
 
+Exclude generated or checked-in output by exact directory component (repeatable):
+
+    cargo run -- build ../yams --exclude site --exclude build \
+      --json yams.brick.json --html yams.brick.html
+
 Layer Fold findings onto the building:
 
     cargo run -- build ../fold \
@@ -22,6 +27,11 @@ Layer Fold findings onto the building:
       --json fold.brick.json \
       --html fold.brick.html
 
+Fold artifact paths are normalized lexically across `/`, `\\`, `.`, and `..`
+components before repository-relative matching. Findings for excluded,
+generated, or genuinely missing files remain diagnostics instead of being
+attached to a neighboring source file.
+
 Render an existing model or enforce an explicit experimental gate:
 
     cargo run -- render fold.brick.json --html fold.brick.html
@@ -53,6 +63,13 @@ multiline signatures. Its score is not proof of correctness, maintainability,
 or security. Use the component metrics and evidence records; do not rank teams
 or repositories by the composite alone.
 
+Unqualified calls such as `lock`, `size`, and `empty` may merge unrelated C++
+targets or remain external; recursion and overloads may also appear as lexical
+self-cycles. File fan-out is therefore a decomposition signal, not proof of
+semantic coupling. Scope architecture gates to production paths and verify each
+candidate against source and tests. Cross-translation-unit and namespaced C++
+resolution remain priority adapter work.
+
 The next serious analyzer milestone should consume compiler/YAMS/Fold graph
 facts through adapters while retaining the standalone lexical lane as a fast
 preview. See docs/architecture/graph-model.md and docs/ROADMAP.md.