A number of papers were published that extended the original BB84 quantum bit commitment protocol. Brassard and Crepéau suggested in 1991 the following protocol to fix the problems with BB84 :
The rationale is that, unlike the earlier protocol, all bases and bits are random, b being encoded in the parity. Alice cannot guess what the decoding bases are, so, would she keep an entangled qubit for herself, measuring it with the wrong basis would yield a 50% probability of error. The authors claimed that this protocol was immune to the EPR attack but were not sure about a higher-scale consequence of quantum physics: ``coherent measurements''. Without explaining much, what they seem to fear is what ultimately would kill the whole enterprise, namely, that the same kind of phenomenon that enabled the EPR attack, the composite state of sets of two or more particles, applies to the composite state of all the particles involved in the protocol.
For a while, however, this was considered to be a sound protocol and the research focused on adding error correction to it, because actual physical quantum transmission channels were beginning to be built and all of them had error probabilities. In 1993 a paper by Brassard, Crépeau, Jozsa and Langlois appeared, pompously named ``A Quantum Bit Commitment Scheme Provably Unbreakable by both Parties''  that explained in full detail an extension to BC91 incorporating error correction.
I won't describe BCJL93 in detail here because the error correction protocol makes equations difficult to follow. This paper contains a proof of security (, theorem 3.7) that turned out to be flawed. Crépeau made this comment three years later (, page 197):
The part of the ``proof'' that goes wrong is the fact that A is committed to a bit. The paper shows that A is unable to know at the same time information to unveil the commitment as b=0 and as b=1 and concludes that A cannot change her mind. The first part of the statement is correct, but not the conclusion. As a matter of fact, the first part of the statement is also true of the BB84 protocol and we know that it can be broken!