The network's admissible demand region (ADR), which is a key index to characterize a network's ability to handle incoming demands, is shaped by each movement's saturation flow rate (SFR). Existing backpressure (BP) traffic control policies commonly assumed a fixed and/or completely known SFR when calculating the pressure for decision-making. On one hand, since real-time traffic conditions can significantly influence the traffic supply, the fixed mean SFR (M-SFR) assumption could result in a mismatch between dynamic demand and supply. On the other hand, accurately predicting the imminent SFR (I-SFR) is challenging because of the complicated interactions between traffic participants. Hence, the completely known SFR assumption is impractical in real-world settings. Our paper demonstrates that, compared with only using the constant M-SFR information, using more knowledge of I-SFR can enlarge the upper bound of ADR. In addition, we theoretically prove that the BP with predicted I-SFR can guarantee network stability as long as the demand is interior to the ADR.The proposed theory is validated by a calibrated simulation model in the experiments. Three I-SFR prediction methods with different accuracies are adopted: the M-SFR method, the heuristic estimation method, and the deep neural network method. They are tested in three BP-based control policies to investigate whether our findings are robust. The simulation results show that: a higher prediction accuracy of I-SFR can effectively help all three BP-based policies enlarge the network ADR, and more accurate I-SFR can productively reduce the average vehicle delay.