According to the n-lab page:
The associative operad Assoc is an operad which is generated by a binary operation $\Theta$ satisfying
$$\Theta\circ(\Theta,1)=\Theta\circ(1,\Theta)$$
It then goes on to say that:
Assoc is hence the operad whose algebras are monoids; i.e. objects equipped with an associative and unital binary operation.
Huh? Where did unitality come from? Sure there's this $1$ floating about, but I thought that just got interpreted as the identity function, as opposed to a distinguished element.
I think something is a bit awry. If you "go back in time" (see previous revisions of the page), you see that first the operad was defined as one whose algebras were associative unital algebras (aka monoids), then someone added that it was generated by a single operation $\Theta$ with the associativity relation.
But if you take the construction of an operad defined by generators and relations (like here), you'll see that if there are no generators of arity $0$, then the resulting operad won't have any operations of arity $0$ (because if $k_i \geq 1$ for all $i$, then $k_1 + \dots + k_r \geq 1$...). So there are two things here:
These two are not equal, of course -- the only difference is in arity $0$.
You are indeed right to say that $1$ is taken to be an identity operation: it's an operation with arity $1$, and in the endomorphism operad it's the identity.