This question is about objects studied in chapter 2 of Lurie's Higher Topos Theory.
Let $p : X \to S$ be a left fibration, and let $e \in X_1$ be an edge of $X$ such that $p(e)$ is an equivalence in $S$ ( - in the sense that the map $p(e) : \Delta^1 \to S$ can be extended along $\Delta^1 \to J$, where $J$ is the groupoid freely generated by $\Delta^1$). Is the edge $e$ $p$-Cartesian?
Proposition 2.4.1.5 implies that this holds true in the case that $S$ is an $\infty$-category.
ADDED
My question can be reformulated succintly to the question of whether the canonical maps $$\phi_n : \Lambda^n_n \sqcup_{\Delta^{\{n-1,n\}}} J \to \Delta^n \sqcup_{\Delta^{\{n-1,n\}}} J$$ are left anodyne, where $n \geq 2$.
Proposition. The map $\phi_2$ is left anodyne.
Proof. Push the left horn $\{0\} \to \Delta^1$ out along the map $\{0\} \to \Lambda^2_2 \sqcup_{\Delta^{\{1,2\}}} J$ that sends $0$ to $0$, and call the new endpoint $1'$, thus obtaining a 1-simplex $(0,1')$. Then fill several inner or left horns, obtaining the following simplices: $(0,1',2), (1',2,1), (1',2,1,2), (0,1',1), (0,1',1,2)$. The map $\phi_2$ is a codomain retract of the resulting map. $\ \blacksquare$
The canonical map $\phi_1 : \{1\} \to J$ is left anodyne, too, of course.
Today I accidentally stumbled upon Corollary 5.2.2.4 of Lurie's Higher Topos Theory:
With this corollary at hand, it is actually quite easy to answer my question:
ADDED