The subgroup $G$ (closure) is made of the permutations on $I_6:=E\sqcup O$ ("even and odd") which send "even-to-even" and (then) "odd-to-odd". Take an $\alpha\in S_6$ which sends some even to odd, but not all the even to the odd, say it "mixes up even and odd"; then $\alpha(E)$ is neither $E$ nor $O$. Accordingly, we can find a permutation $\sigma\in G$ which sends $\alpha(E)$ to a set distinct from $\alpha(E)$: for example, legitimate such $\alpha$'s and $\sigma$'s are all those such that:
$$\alpha(2)=\color{red}{1}, \space\space\space\alpha(4)=\color{blue}{4}, \space\space\space\alpha(6)=\color{green}{2}\tag 1$$
and
$$\sigma(\color{red}{1})\space\stackrel{odd-to-odd}{=}\space3, \space\space\space\space\sigma(\color{blue}{4})\space\stackrel{even-to-even}{=}\space2, \space\space\space\space\sigma(\color{green}{2})\space\stackrel{even-to-even}{=}\space4$$
So, by choosing "mixing up" $\alpha$'s ("exists..."), we have been able to build up $\sigma$'s ("exists...") such that:
$$\sigma(\alpha(E))=\{3,2,4\} \ne \{1,4,2\}=\alpha(E)=\alpha(\tau(E)), \space\forall \tau\in G \tag 2$$
We can read $(2)$ this way:
$$\exists\alpha\in S_6, \exists \sigma\in G\mid (\sigma\alpha)(E)\ne (\alpha\tau)(E), \space\forall \tau \in G \tag 3$$
whence:
$$\exists\alpha\in S_6, \exists \sigma\in G\mid \alpha^{-1}\sigma\alpha\ne \tau, \space\forall \tau \in G \tag 4$$
or equivalently:
$$\exists\alpha\in S_6, \exists \sigma\in G\mid \alpha^{-1}\sigma\alpha\notin G \tag 5$$
meaning $G\ntrianglelefteq S_6$.