It’s purpose is to end up with a single IRR where ordinary IRR calculations would give several IRRs. It also assumes that cash receipts are reinvested at the discount rate rather than the IRR (an assumption implicit in ordinary IRR).
Personally, I think it’s all a bit silly because it doesn’t tell you anything about project acceptance rejection that a simple NPV calculation wouldn’t. Like ordinary IRR it does not help you decide between mutually exclusive projects of different sizes. Thare again NPV does the job much better.