I have been looking into PRNGs and stumbled upon the efficient Jump Ahead implementation for Multiplicative Lagged Fibonacci Generator. The technique relies on the fact that any odd integer $X$ can be uniquely represented modulo $2^m$ as $x \equiv (-1)^y 3^z \mod 2^m$, where $y \in \{0,\ 1\}$ and $z \in \{0,\ \ldots,\ 2^{m-2}-1\}$. I find myself unable to prove this fact or find anything on the Internet.
I would greatly appreciate it if someone could help me proving this.
For any $n\geq 3$ we have $$\mathbb{Z}/(2^n \mathbb{Z})^* \simeq \mathbb{Z}/(2^{n-2} \mathbb{Z})\rtimes\mathbb{Z}/(2\mathbb{Z})$$ and every element of $\mathbb{Z}/(2^n \mathbb{Z})^*$ can be represented as $\pm 5^h$ since $5\equiv 1\pmod{4}$, $5$ generates the quadratic residues in $\mathbb{Z}/(8\mathbb{Z})^*$ and we may apply the Hensel lifting lemma. We have something similar (namely, your statement) if $5$ is replaced by $3$.