I'd like to prove that we can get rid of Conditions 2 and 4 from your listNot an answer, but a question based on a previous attempt of wishesfinding a solution, for then,which Benjamin Steinberg pointed out a foolish mistake in order to conclude, it will be enough to consider your preferred finite commutative band and extend it, if necessary, according to the stepscomments below.
For, suppose that we have somehow built Is there an "enlarge-and-shrink recipe" to extend a semigroup $\mathbb A = (A, \cdot)$, either finite or infinite, for which Conditions 1 and, 3 and 4 hold true. First, we unitize $\mathbb A$ by forcing an identity even if $\mathbb A$ is already unital. It is seen that the outcome is still to a larger semigroup, let me denote it by $\mathbb A^{(1)}$$(S, \cdot)$ for which Conditions 1 and, 3 and 4 continue to be true (unless I'm missing something). More interestingly, we have a gainbut in the process, because both of the regular representations of $\mathbb A^{(1)}$ are now faithfuladdition (in the sense of your definition). This shows that we can assume from$S^2 \ne S$? If the outset that $\mathbb A$answer is a monoid fulfilling Conditions 1yes, 3 and 4 (as forthen the latter, evenproblem is solved in a stronger form).
At this point, pick an element $\infty \notin A$ and extend $A$ to the set $S := A \cup \{\infty\}$ and $\cdot$ to a binary operation on $S$, here still denoted bypositive: Start with your preferred self-dual band $\cdot$$\mathbb A$, unitize it by taking $\infty \cdot \infty := e$ and $\infty \cdot x := x \cdot \infty := x$ for $x \in S$, where $e$ is theadjoining an identity ofonly if $\mathbb A$. Then, $(S, \cdot)$ is still a semigroup for which Conditions 1, 3 and 4 are verifiednot already unital (again, unless I'm missing something), but in addition $S \cdot S = A \subsetneq S$.
Added later. As forin such a concrete example (based onway that the above)unitization is still a self-dual band, but we can just considerhave a set $S$ with three elements $1$, $e$ and $\infty$ and let $\cdot$ begain in the binary operation on $S$ given byprocess, since now the following (Cayley) table:
$$\begin{array}{c|ccc} \cdot & 1 & e & \infty \\ \hline 1 & 1 & e & 1 \\ e & e & e & e \\ \infty & 1 & e & e \end{array}.$$ The recipeoutcome is: We start with a semigroup whose regular representations are both faithful, regardless as to whether or not this was already the trivial groupcase with carrier $\{1\}$, then we unitize it by forcing a new identity $e \ne 1$$\mathbb A$), and lastly we adjoin a further element $\infty \notin \{1, e\}$ and extendfinally use the multiplication table by taking $\infty \cdot \infty := e$ and $\infty \cdot x = x \cdot \infty = x$ for $x = 1$ or $e$enlarge-and-shrink recipe to conclude.