For solar cells with absorbers made of co-evaporated Cu(In,Ga)Se-2 (CIGS), the two-stage process originating from Boeing is a familiar recipe. A key issue for this process is that the film during the first stage is grown Cu-rich, after which the composition in the second stage evolves to be Cu-poor. In our extreme version of the Boeing recipe, which we denote CURO, the second stage is performed completely without Cu flux. The CURO films have relatively large grains, but a rough surface with crevices between the grains. In this paper we examine a reversed two-stage process, INRO, where the metal fluxes are constant in the fist stage and yield a Cu-poor film. In the second stage, the In and the Ga sources are turned off, and the film evolves gradually less Cu-poor until the final composition is reached. For both processes, endpoint detection (EPD) is used to achieve a predetermined composition of the CIGS. Comparing these films with the ones from CURO, the grains from the reversed process are smaller and the top surface is smoother. This is true for CIGS films with Cu-poor stoichiometry with a Cu content sufficiently low to fabricate solar cell devices. If the process continues, the grains undergo a transformation and grow to grain sizes usually seen in CIGS films grown under Cu-rich conditions and, finally, the stoichiometry becomes Cu-rich. The crevices from the CURO process have been explained by transport of Cu from CuxSe, which during the first stage is segregated in the grain boundaries and subsequently consumed during the second stage. For the reversed process, no similar material transport is observed. Devices that originated from the INRO process did not perform well, but the process is an interesting key for understanding the grain restructuring observed close to Cu-poor/Cu-rich transitions. (c) 2005 Elsevier B.V. All rights reserved.