This starts with a vaguely-recalled result (which may be false!): that if $\mathcal{U}$ is a measure on the least measurable cardinal $\kappa$, then every elementary $j: L[\mathcal{U}]\rightarrow M$ such that $(j,M)$ is first-order definable over $L[\mathcal{U}]$ with parameters from $L[\mathcal{U}]$ "comes from" iterating the elementary embedding associated to $\mathcal{U}$ itself. That's a bit vague of course, but it just occurred to me that I don't know how to whip up a situation where the opposite extreme holds:
Is it consistent with $\mathsf{MK}$ (relative to large cardinals) that for each $n$ there is a definable pair $(j,M)$ such that $j:V\rightarrow M$ is elementary but $(j,M)$ is not $\Sigma_n$-definable?
Here "definable" means "first-order definable over $V$ with set parameters." Both $j$ and $M$ are proper-class-sized objects ($M$ literally is a proper class, $j$ is a class function) so this is nontrivial; correspondingly, I'm using $\mathsf{MK}$ as my base theory here so that I can directly phrase the principle in question.
I strongly suspect that the answer to the question is yes but I don't know how to produce a model in which this holds.