If G is a Lie group acting on a manifold M through $\Psi$, one can argue that orbit of any $p\in M$ is an integral submanifold of the generators of group action. Roughly the proof is :
1) Fixing p first one proves that the orbit is a submanifold through following commuting diagram:
$G_p$ being the isotropy group of p. $\pi$ is a sumbersion and $\widetilde{\Psi^p}$ a immersion. Thus the image of $\mathfrak{g}$ under $d\Psi^p$ covers the tangent space of $O_p$.
2)Also \begin{equation} d\Psi^p(X_g)=d\Psi^p\circ dL_g(X_e)=d(\Psi^p\circ L_g)(X_e)= d\Psi^{g.p}(X_e)= \frac{d}{d\epsilon}\Big|_0\Psi(exp(\epsilon\textbf{v}),x) \end{equation}
The curve $\Psi(exp(\epsilon\textbf{v}),x)$ is in the orbit, and hence $d\Psi^p(X_g)$ is in the tangent space of the orbit.
The above two facts establish that orbit is integral manifold of generators of group action.
My question is when can one say that the orbits are maximal integral manifolds.
I have a feeling that oribits will be maximal integral manifolds if G is connected else the integral manifold through p will be a subset of the orbit of p.
Your assertion comes the fact that $M$ is the disjoint union of maximal integral manifolds and of orbits. A maxiam integral manifold $N$ satisfies the following : if $p\in N$ its orbit is also in $N$, as an open sub-manifold, (open for the manifold topology). So $N$ is the disjoint union of orbits, these orbits are open. By connectedness $N$ is an orbit.