I agree that it is difficult to see how $f:=\prod_i(1+x_i)$ expands into homogenous poylnomials. But it is still easy to see why a contradiction arises.
If we were to expand $f$ into elementary symmetric functions, the result would be the infinite sum
$$
f=1+e_1+e_2+\dots
$$
This is the extension of the finite equality $$(1+x_1)(1+x_2)(1+x_3)=1+(x_1+x_2+x_3)+(x_1x_2+x_1x_3+x_2x_3)+x_1x_2x_3.$$
Now, applying the $\omega$ involution to both sides,
$$
\omega(f)=1+h_1+h_2+\dots
$$
Next, expand $\omega(f)$ into the $m$ basis. The result is
$$
\omega(f)=\sum_{\lambda }m_\lambda,
$$
since each $m_\lambda$ appears in exactly one $h_n$. Finally, we can see that (using the convention $h_{(0)}=m_{(0)}=1$ for convenience)
$$
\langle \omega(f),\omega(f)\rangle=\left\langle \sum_{i=0}^\infty h_{(i)},\sum_\lambda m_\lambda\right\rangle=\sum_{i=0}^\infty \langle h_{(i)},m_{(i)}\rangle = \sum_{i=0}^\infty 1=\infty
$$
We do not want infinite values of the inner product. It may not seem like problem, but if $\infty$ can arise, then so can $\infty-\infty$, which is a problem. For example, computing the inner product of $\sum_{i=1}^\infty (-1)^i h_{(i)}$ and $\sum_{\lambda}m_\lambda$ would be $1-1+1-1+\dots$, which is undefinable.