Unit tests in object-oriented programming languages must instantiate objects as an essential part of their set-up. Finding feasible method-call sequences for object creation and selecting a most desirable sequence can be a time-consuming challenge for developers in large C++ projects. This is caused by the intricacies of the C++ language, complexity of recursive object creation, and a large number of alternatives. We confirm the significance of the problem by analysis of 7 large C++ projects and a survey with 143 practitioners. We then design an approach for recommending method-call sequences for object creation that align with criteria gathered by the survey. Our approach exploits accurate and efficient compiler-based source code analysis to build an object dependency graph that is processed by a divide-and-conquer algorithm.An evaluation on a large industrial project shows that our tool finds solutions that require in 99% of 1104 cases identical or fewer objects compared to manually crafted solutions. Developer feedback and manual analysis confirm these results. Moreover, solutions found by our tool require up to 6 times fewer objects on average compared to approaches from prior work.