In Advanced Microeconomic theory by Jehle and Reny there is a proof of the theorem which states the existence of utility function.
In order to prove the existence of a utility function $u(\mathbf{x})$ which represents binary relation $\succeq$ if it is complete, transitive, continuous, and strictly monotonic, it is suggested to consider a mapping
$u: \mathbb{R_+^n} \to \mathbb{R}$ such that $u(\mathbf{x})e\sim \mathbf{x}$ is satisfied, where $\mathbf{x}$ is a bundle, $u(\mathbf{x})$ is some number and $\mathbf{e}$ is a bundle which contains one of every good.
So first we need to show that there always exists such number $u(\mathbf{x})$. To do this consider two sets:
$A \equiv \{t \geq 0 \mid t\mathbf{e} \succeq \mathbf{x}\}$
$B \equiv \{t \geq 0 \mid t\mathbf{e} \preceq \mathbf{x}\}$
if $t^* \in A \cap B$, then $t^*\mathbf{e} \sim \mathbf{x}$, so we need to show that $A \cap B$ is nonempty.
The continuity of $\succeq$ implies that both A and B are closed in $\mathbb{R_+}$. By strict monotonicity, $t \in A$ implies $t' \in A,$ $\forall$ $ t'\geq t$. So $A=[\underline{t}, \infty)$. Similarly $B=[0, \overline{t}]$
For any $t \geq 0$, completeness of $\succeq$ implies that either $t\mathbf{e} \succeq \mathbf{x}$ or $t\mathbf{e} \preceq \mathbf{x}$, i.e., $t \in A \cup B$
$\mathbb{R_+} = A \cup B = [0, \overline{t}] \cup [\underline{t}, \infty)$.
So in order to $A \cap B$ be nonempty it should be that $\underline{t} \leq \overline{t}$.
But is it always like this? I can't see why the last inequality should hold always.