A new algorithm is described that predicts branching architectures of polymers from radical polymerization with transfer to polymer and termination by disproportionation only in a continuously stirred tank reactor (CSTR). An existing full Monte Carlo (MC) method(1-3) generates architectures of molecules with varying overall numbers of monomer units (or chain length), n, and numbers of branch points, N. The new algorithm is a conditional MC method that generates molecules of given specific n and N. Both methods employ primary polymers as the linear building blocks of branched molecules. The time order of primary polymer coupling, being determined by the residence time distribution of primary polymers, turns out to be crucial for the architectures. A new and elegant recursive scheme rigorously calculates the primary polymer lengths in coupling order directly from the overall chain length/number of branch points distributions (CLD/DBD), P(n,N), without explicitly dealing with primary polymer residence times. This proved assumptions made in earlier algorithms(4,5) to be erroneous. Primary polymer length and branching density distributions as well as radius of gyration contraction factors from full and conditional MC show excellent agreement. Statistically representative samples of architectures for given n and N are thus generated in a fast manner. As an application example, a comparison is made to the contraction factor as predicted by an old relation by Zimm and Stockmeyer.(6) It is concluded that the heterogeneous branching of the present chemical case leads to stronger contraction than the homogeneous branching case described by the latter authors.