$\textbf{Fact 1}$ In principle, the QCD Lagrangian should contain a Lorentz invariant, gauge invariant, dimension-4 term $\sim\theta \text{Tr}[F^{\mu\nu}\tilde{F}_{\mu\nu}]$. This term, however, is usually neglected in classical physics because it is a total divergence, and therefore, cannot affect the equations of motion. However, when instanton effects are included, it turns out that the QCD action should be augmented by this term. This is explained, for example, in the book Quantum Field Theory by Mark Srednicki at the pages 598-599.
$\textbf{Fact 2}$ The part of the QCD Lagrangian with $u$ and $d$ quark, in the massless limit of $u$ and $d$, has a $U(1)$ axial anomaly $$\partial_\mu j^{\mu 5}=-\frac{g^2}{16\pi^2}\epsilon^{\alpha\beta\mu\nu}F_{\alpha\beta}^a F_{\mu\nu}^a\sim \text{Tr}[F\tilde{F}]$$ where $j^{\mu 5}=\bar{Q}\gamma^\mu\gamma^5 Q$ and $Q=\begin{pmatrix}u\\d\end{pmatrix}$. This anomaly term is exactly same in form as the term included in the QCD action by instanton effects.
Question This uncanny similarity strongly provokes me to guess that this anomaly is solely responsible for inducing the $\theta-$term $\sim\theta \text{Tr}[F^{\mu\nu}\tilde{F}_{\mu\nu}]$ to the QCD action. In other words, if the current $j^{\mu 5}$ were conserved or anomaly-free i.e., $\partial_\mu j^{\mu 5}=0$, the term $\sim \theta \text{Tr}[F^{\mu\nu}\tilde{F}_{\mu\nu}]$ to QCD action can always be dropped and instanton effects will not be present.
Is this the correct way to think about this striking correlation between fact 1 and fact 2? I'm doubtful because while discussing instantons of Yang-Mills action, in pages 590-599, Srednicki doesn't talk about anomalies at all.