Attribution debates usually go in circles because people argue about which model is “right” instead of which model fits the decision in front of them.
Match the model to the decision
Last-click is fine for daily bid adjustments. Multi-touch is better for budget allocation across channels. Incrementality testing is the only honest answer for “should this channel exist at all.”
The mistake I see most
Using last-click data to justify a channel’s existence, when the question it was built to answer is much narrower than that.