The limits of maximizing the open-circuit voltage Voc in solar cells based on poly[2,7-(9,9-didecylfluorene)-alt-5,5-(4,7-di-2-thienyl-2,1,3-benzothiadiazole)] (PF10TBT) as a donor using different fullerene derivatives as acceptor are investigated. Bulk heterojunction solar cells with PF10TBT and [6,6]-phenyl-C61-butyric acid methyl ester (PCBM) give a Voc over 1 V and a power conversion efficiency of 4.2%. Devices in which PF10TBT is blended with fullerene bisadduct derivatives give an even higher Voc, but also a strong decrease in short circuit current (Jsc). The higher Voc is attributed to the higher LUMO of the acceptors in comparison to PCBM. By investigating the photophysics of PF10TBT:fullerene blends using near-IR photo- and electroluminescence, time-resolved photoluminescence, and photoinduced absorption we find that the charge transfer (CT) state is not formed efficiently when using fullerene bisadducts. Hence, engineering acceptor materials with a LUMO level that is as high as possible can increase Voc, but will only provide a higher power conversion efficiency, when the quantum efficiency for charge transfer is preserved. To quantify this, we determine the CT energy (ECT) and optical band gap (Eg), defined as the lowest first singlet state energy ES1 of either the donor or acceptor, for each of the blends and find a clear correlation between the free energy for photoinduced electron transfer and Jsc. We find that Eg - qVoc > 0.6 eV is a simple, but general criterion for efficient charge generation in donor-acceptor blends.