1) You have a typo at
"Focusing on clean will at best leave you with bad code. Focusing on delivering business value will leave to at worst good code.."
Good business value leads to two outcomes: Excellent code or clean code.
Clean code leads to two outcomes: Excellent code or bad code.
It should be
"Focusing either on clean code or high business value can lead to excellent code. However, the worst case scenario on high business value is good code while the worst case scenario on clean code is bad code"
2) I *love* the DIRTI process and that you break it up into 2 separate loops. DIR to address the lack of understanding of the problem space and therefore increase chances of high business value. RTI to address the code cleanliness issues.
I only want to add that Dan North has a presentation at infoq where he argues the same weakness of TDD like you did here. He spoke of how he saw highly productive teams doing the spike-and-stabilize. They wrote code first as prototype and they were willing to throw the prototype away and redo from scratch. This is the spike -- similar to your DIR.
The stabilize is where the unit tests and TDD stuff comes in. Analogous to your RTI.
Your DIR loop can be strengthened by allowing the possibility of writing throw-away prototypes.
Perhaps a DIR-T where T stands for throw away so you can develop from scratch with better understanding.
In my personal experience, T happens quite often precisely because tackling an unfamiliar problem space leads more often than not poor assumptions and these poor assumptions may get accidentally built into your initial architectural design.
Nothing less than a rewrite can remove these damaging assumptions.
Of course, I don't mean every iteration of the DIR loop must end with a throwaway. But due to behavioral psychology phenomena like sunk cost fallacy and the Endowment effect, T should be explicitly be made known as a possible tactic in the early DIR cycle.
You (and many people like you) inspire me to put down in words my thoughts as well. I intend to build a similar body of written work like yours.