STRLEN is currently documented:

The strlen function corresponds to the XPath fn:string-length function and returns an xsd:integer equal to the length in characters of the lexical form of the literal.

fn:string-length is defined as "[returning] the number of ·characters· in a string," with character being "an instance of the Char production" from XML, which is defined as:

[2]   	Char	   ::=   	#x9 | #xA | #xD | [#x20-#xD7FF] | [#xE000-#xFFFD] | [#x10000-#x10FFFF]

Since RDF strings are defined as sequences of Unicode scalar values, I think there's a gap here regarding the code points below U+0020 (modulo the 3 explicitly enumerated in the production). This gap should be observable in considering this query:

SELECT (STRLEN("\u0001\u0002") AS ?l) {}

I think a plain reading of F&O suggests that this would return zero, while the intuitive (IMO) answer is two.

Perhaps we need to update the definition of STRLEN to use language that does not bottom-out in XML characters, but instead talks directly about Unicode code points and/or scalar values.