That's a misrepresentation of Alex Cowan's work.

Cowan suggests starting with protopersonas - assumptions about what we believe users and customers need and how they behave.

He then suggests testing those protopersonas through interviews, which are then revised into validated qualitative personas.

From these interview transcripts, he recommends creating problem charts: what the user is trying to get done, their current solutions, and the characteristics a solution would have to have in order to develop something that is better than the current solution.

The idea is then generated based on the observations you collected during your research.

While I don't disagree with your argument, it is not an accurate presentation of Cowan's continuous-research-based framework.