Recently, a quantum secure direct communication (QSDC) protocol based on the mixture of Bell state particles and single photons[Acta Phys. Sin. 65 230301(2016)] was put forward. In this QSDC protocol, the single photons and the Bell states were both used as information carriers. To be specific, each Bell state as well as single photon was encoded by three bits of classical information. After the sender told the receiver how to measure the particles, the receiver could read out the secret message sent by the sender. Speciously, the information transmission efficiency of this protocol was high. Unfortunately, there exists the information leakage problem in this protocol. When the sender announces that the receiver uses the Z-basis to measure a single photon, everyone knows that the sent secret message is 000 or 001, that is, the first two bits are leaked out; when the sender announces that the receiver uses the X-basis to measure a single photon, everyone knows that the sent secret message is 010 or 011, that is, the first two bits are leaked out too; when the sender announces that the receiver uses the Bell-basis to measure a pair of particles from a Bell state, everyone knows that the sent secret message is 100, 101, 110 or 111, that is, the first bit is leaked out. In a word, two of the three bits of classical information encoded in a single photon, and one of the three bits of classical information encoded in a Bell state are leaked out. Therefore, this scheme is not secure. On the basis of keeping the original idea and changing the contents of the protocol as less as possible, we put forward an improved message encoding rule to solve the information leakage problem, that is, the single photon is only encoded by one bit of classical information, and the Bell state is only encoded by two bits of classical information. In fact, this makes the information capacity of the improved protocol achieves the Helovo bound. So it has high coding capacity. We hope researchers pay more attention to the information leakage problem in quantum secure communication protocols, and thus design truly secure ones.