No, you're confusing expressions with metanotation.
Using infix metanotation to describe Lisp semantics is a convention that goes back to McCarthy.
It's pretty universal to use syntax that's outside the domain of the language being described to describe a language in it's specification, otherwise your head would explode trying to distinguish the example from the rule.
The C++ standard uses tables for the equivalent purpose, not C++ declarations, like at §23.2.1 in C++11. Does this mean that C++ should move from header files to an RDBMS?
(car nil) => NIL
Shouldn't they say, for example:
(eq (car x) (first x))
Sure, it's the same, both imply a test for equality or in this case an assertion of equality. However, '==' is not part of this language and there is reason to use it other than the alternative is less obvious and clear.