LATE rarely captures a policy-relevant treatent effect in the first place, so why should be care about the difference between IV and LATE?I disagree with the lack of relevance of the LATE.
IV is an estimator and LATE a parameter. IV will estimate the LATE under some assumptions. So your sentence makes no sense.
If the IV estimator estimates a weighted average treatment effect with positive and negative weights, then it can give meaningless results.
But from my point of view, this paper is making an obvious point.
No, LATE is not a parameter to an economist. Stop reading stats articles and revisit your core sequence.