It is easy to prove that $\omega_1$ does not have an order-preserving injection into $(\mathbb R, <)$: to each $\alpha \in \omega_1$ we could assign the interval between the image of $\alpha$ and $\alpha + 1$, and this gives uncountably many disjoint open intervals in $\mathbb R$, contradicting its separability.
Intuitively it seems to be that if $\alpha < \omega_1$, there should be an order-preserving injection into $\mathbb R$, but I am having trouble constructing it. If we have $\alpha \to \mathbb R$, we can compose it with a homeomorphism $\mathbb R \to (0, 1)$, and find $\alpha + 1 \to \mathbb R$ by sending $\alpha$ to $(0, 1)$ and $\{\alpha\}$ to 2. However, the limit step is more difficult: if $\lambda < \omega_1$ is a limit, then it is a limit of countably many well-orderings which inject in an order-preserving manner into $\mathbb R$, but you cannot expect these to be "compatible" in a way obvious to me.
Is there a way to make this (or another argument) work? Or are there countable ordinals with no order-preserving injections into $\mathbb R$?