The following question is taken from Arrows, Structures and Functors the categorical imperative by Arbib and Manes.
Background information
Prove that if $\phi:G\rightarrow H$ is a group homomorphism with $\phi(G)=M\neq H,$ then $\phi$ is not an epimorphism in $\textbf{Grp}.$ [Hint: If $M$ has only 2 cosets in $H$, take $K=H/M,$ and define $\psi, \psi':H\rightarrow K$ by $\psi(h)=[h]$ and $\psi'(h)=e_K.$ If not, take $K$ to be the group of all permutations of $H.$ Choose $3$ different cosets $M, Mh', Mh''$ of $M$ in $H,$ and define $\sigma$ in $K$ by $\sigma(xh'')=xh', \sigma(xh')=xh''$ for $x\in M,$ while otherwise $\sigma(h)=h.$ Define $\psi,\psi':H\rightarrow K$ by $\psi(h)=$ left multiplication by $h$, and $\psi'(h)=\sigma^{-1}\psi(h)\sigma.$ Then in either case $\psi\phi=\psi'\phi$ but $\psi\neq \psi'.$]
Questions
I don't understand how $\psi \neq \psi'.$
According to the proof, suppose $h\in H\setminus M,$ then $\psi(xh'')=hxh'',$ and $\psi'(xh'')=(\sigma^{-1}\psi\sigma)(xh'')=\sigma^{-1}\psi\sigma(xh'')=\sigma^{-1}\psi(xh')=\sigma^{-1}(hxh'')$
But does $\sigma^{-1}(hxh'')=\sigma^{-1}(h)\sigma^{-1}(xh'')?$ (not sure if this is correct or how to justify it?) So that $\sigma^{-1}(h)=h$ and $\sigma^{-1}(xh'')=xh'.$ Hence $\sigma^{-1}(hxh'')=hxh'.$ Hence $\psi'(xh'')\neq \psi(xh'').$
Am I understanding the proof correctly? Thank you in advance