It seems there is a typo in the article. It likely meant instead something like that if there are no (rather than "a") non-trivial $x$ satisfying the specified congruence equation, then $2^p-1$ is prime. This is also suggested in the article text where, between its equations $(5)$ and $(6)$, it states that
So, if there exists any integer lattice point on the hyperbola, $M_p$
is not prime.
There's a fairly simple and direct way to show this compared to what's used in the linked article. Since $\gcd(2, 2x + 1) = 1$ (so $2^{-1} \equiv x + 1 \pmod{2x + 1}$), we have
$$\begin{equation}\begin{aligned}
2^{p-1} + x & \equiv 0 \pmod{2x+1} \iff \\
2^{p} + 2x & \equiv 0 \pmod{2x+1} \iff \\
2^{p} - 1 & \equiv 0 \pmod{2x+1}
\end{aligned}\end{equation}$$
If $2^{p} - 1$ is prime, then the only positive values of $2x + 1$ which work are $2x + 1 = 1 \;\to\; x = 0$ and $2x + 1 = 2^{p} - 1 \;\to\; x = 2^{p-1} - 1$, i.e., what are considered to be the trivial solutions. Thus, if there are any non-trivial solutions, then $2^p - 1$ cannot be prime, i.e., it's composite as you've determined. In particular, each factor of $2^p - 1$, with $1 \lt f \lt 2^p - 1$, has a solution of $f = 2x + 1$ (e.g., for $p = 4$ with $2^4 - 1 = 15$, we have the solutions of $2x + 1 = 3$ (so $x = 1$) and $2x + 1 = 5$ (so $x = 2$)).