Nonorthogonal multiple access (NOMA) promises higher throughput than the conventional orthogonal access by multiplexing the signal of a weak user with that of a strong one. Identifying the right combination of users with an appropriate assignment of power appears to be the bottleneck in achieving optimal throughput in NOMA-driven systems. Despite the large volume of literature, it is still not well understood how the task of joint user selection and power allocation (PA) can be efficiently performed in multiband NOMA systems in the presence of a large number of users'. To address this task, we propose a variable PA (VPA) scheme across sub-bands and evaluate its performance. Contrary to previous studies, we find that the proposed VPA scheme significantly outperforms the previously known schemes when the number of users and variation in their associated signal-to-interference-plus-noise ratios (SINRs) are large. Our investigation also helps to visualize how the system throughput, the average number of selected users, and the system fairness change as functions of system parameters like SINRs and the number of system users.