Feel free to suggest a better title.
We're going to regard $0$ as a limit ordinal, as people sometimes do. Let $L$ be the set of countable limit ordinals.
If $\alpha\in\omega_1$ there exist a unique $\beta_0\in L$ and $n_0\in\omega$ such that $$\alpha=\beta_0+n_0.$$
Now $L$ is evidently an uncountable well-ordered set such that every element has at most countably many predecessors, so there exists an order isomorphism $$\phi:\omega_1\to L.$$
So $\beta_0$ is $\phi$ of something, hence there exist a unique $\beta_1\in L$ and $n_1\in\omega$ with $$\beta_0=\phi(\beta_1+n_1).$$And so on. We've defined a map from $n:\omega_1\to\omega^\omega$ (where by $\omega^\omega$ I mean the set of all maps from $\omega$ to itself): $$n(\alpha)=(n_0,n_1,\dots).$$
Question: Is $n:\omega_1\to\omega^\omega$ injective?
(If yes then this gives a more elegant construction of sets serving the purpose of the $I_{n,j}\subset\omega_1$ here.)
It seems to be clear that $\phi$ is continuous, so we can construct a nonzero fixpoint for it in the usual way:
$$ \alpha = \bigcup_{k\in\omega} \phi^k(1) $$ Clearly $\alpha\le\omega_1$, but since it has cofinality $\omega$ it can't be $\omega_1$ itself. Thus $\alpha < \omega_1$ and $\phi(\alpha)=\alpha$.
Now $n(\alpha)=(0,0,0,\ldots) = n(0)$, so $n$ is not injective.