Finish up pattern-matching.md

Author: Sporarum

docs/_docs/reference/changed-features/pattern-matching.md

@@ -63,14 +63,16 @@ and `S` conforms to one of the following matches:
 The former form of `unapply` has higher precedence, and _single match_ has higher
 precedence over _name-based match_.
 
+**Note:** the `S` in `R` can be `U`.
+
 A usage of a fixed-arity extractor is irrefutable if one of the following condition holds:
 
 - `U = true`
 - the extractor is used as a product match
 - `U <: R` and `U <: { def isEmpty: false }`
-- `U = Some[T]` (since `Some` has `def isEmpty: Boolean` for compatibility reasons)
+- `U = Some[T]`
 
-**Note:** the `S` in `R` can be `U`.
+**Note:** The last rule is necessary because, for compatibility reasons, `isEmpty` on `Some` has return type `Boolean` rather than `false`, even though it always returns `false`.
 
 ### Boolean Match
 
@@ -201,7 +203,9 @@ A usage of a variadic extractor is irrefutable if one of the following condition
 
 - the extractor is used directly as a sequence match or product-sequence match
 - `U <: R` and `U <: { def isEmpty: false }`
-- `U = Some[T]` (since `Some` has `def isEmpty: Boolean` for compatibility reasons)
+- `U = Some[T]`
+
+**Note:** The last rule is necessary because, for compatibility reasons, `isEmpty` on `Some` has return type `Boolean` rather than `false`, even though it always returns `false`.
 
 **Note:** Be careful, by the first condition and the note above, it is possible to define an irrefutable extractor with a `def isEmpty: true`. 
 This is strongly discouraged and, if found in the wild, is almost certainly a bug.