The classification came first. As Killing says in his introduction (translated by Coleman (1989)):
For each $l$ there are four structures supplemented for $l \in \{2, 4, 6, 7, 8\}$ by exceptional simple groups. For these exceptional groups I have various results that are not in fully developed form; I hope later to be able to exhibit these groups in simple form and therefore am not communicating the representations for them that have been found so far.
More details in Hawkins’s work (quoted in your linked reference) (1982, p. 156):
The discovery that type $G_2$ actually exists seems to have transformed Killing’s attitude towards the possible existence of further new simple groups. (...) Indeed, whereas, initially, Killing’s errors of calculation had convinced him that no new simple groups exist, after this discovery, when he obtained the $\smash{a_{ij}}$ and associated root systems for the exceptional types $E$ and $F$ and the “new” general type $\smash{C_l}$, Killing maintained that simple groups for these types exist even though he never managed to carry his calculations far enough to be able to write down multiplication tables for them.
and in Cartan (1894, pp. 94-95):
As to the determination of simple group structures, Mr. Killing is content to show that to types A), B), D) there correspond long known simple groups, namely the general projective group of $l$-dimensional space and the projective groups of a nondegenerate quadric in spaces of $2l$ and $2l-1$ dimensions. But he doesn’t give the explicit calculations allowing to show that they are the only ones, and as to the other integer systems, he only gives vague indications on the corresponding structures, except however for type C).