Expressions like $X castable as xs:NOTATION

The expression "$X castable as xs:NOTATION" is supposed to return a static error, XPST0080. I was wondering why this is, as the type xs:NOTATION is a valid type name. I understand that the error should be thrown with "$X cast as xs:NOTATION" because there is no constructor function for xs:NOTATION, but surely, just *testing* whether it is castable using "castable as" should never raise an error?

It is different for xs:anySimpleType and xs:anyAtomicType because these type names are not available at all.

Furthermore, we say in FO31, section 18.2:

"There is no constructor function for xs:NOTATION. Constructors are defined, however, for xs:QName, for types derived or constructed from xs:QName, and for types derived or constructed from xs:NOTATION."

The last part of this sentence seems untrue "or constructed from xs:NOTATION". How is this possible if there is no constructor?

And I'm puzzled by the general rule that you can cast from specific to generic. I.e., a type derived from xs:decimal (like xs:positiveInteger) can always be cast to xs:decimal. But here we seem to imply that a type derived from xs:NOTATION can never be cast to xs:NOTATION, even though the inheritance hierarchy would suggest we could.

Note, I am not raising this as a bug, as I am quite sure this topic has received ample attention already in the XP 2.0 times, I am just wondering why these discrepancies exist.

Thanks,
Abel