Hence my point about using measurements from historic actuals to give a forecast instead of asking for explicit estimates. Sitting some devs down and asking them for gut feels is far less accurate than comparing to previous increments.

Certainly! "Gut feel" is a shit way to do time estimates. It's a programmer way, not an engineering way. Of course it's harder to get historical data for software than for other engineering fields, because software is so new. But it's not impossible, and if companies don't even collect the data (due to using story points) they'll never manage to get forecasts when they need them.