In that On-semi article the leakage all referred to the primary is assumed to be Lr, this is a fairly close approximation, as is assuming you can measure Lp ( Lprimary ) with the sec open - actually this measurement is Lpri ( actually Lmag ) + L'leak, where L'leak is that leakage associated with the pri that does not couple to the sec.
But these approximations get you near the ball park. [ just for completeness the measurement when sec is shorted is L'leak + L"leak, where L"leak is the sec leakage ( that which does not couple to the pri ) reflected to the pri side, there is also the parallel ( shunt ) path of Lmag, to some extent L'leak and Lmag cancel each other to get closer to L'leak ( ~ = Lr ) ].