TL;DR: The main point is that boundary conditions (BCs) are important, but they manifest themselves in different ways for bound and scattering states.
Fact: Recall that if we ignore BCs, then for a given arbitrary energy $E\in\mathbb{R}$ (of either sign!), the 1D TISE$^1$ is a 2nd-order homogeneous ODE. Its 2 linearly independent solutions can always be chosen to be both real. If furthermore the potential $V(x)=V(-x)$ is even, then the 2 linearly independent solutions can always be chosen to be even and odd, respectively, cf. e.g. this and this Phys.SE posts.
Bound states $E<0$. If a solution $\psi(x)$ is a bounded function $\mathbb{R}\to \mathbb{C}$, it automatically vanishes at $x=\pm\infty$, so the BCs $\lim_{|x|\to \infty}\psi(x)=0$ are automatically satisfied. Not surprisingly, the above fact continuous to hold: We can choose $\psi$ to be real (and even/odd if the potential $V(x)=V(-x)$ is even).
Scattering states $E>0$. There are generically no reasons why the incoming asymptotic left- & right-moving waves in the 1D scattering problem should display an even/odd symmetry, even if the potential $V(x)=V(-x)$ is even.
--
$^1$We will in this answer assume that the potential vanishes asymptotically: $\lim_{|x|\to \infty}V(x)=0$.