I've helped another user that asked me me in private exactly the same issue as yourself and my solution worked for him... you may wanna try it as well:

Let me know if it works so I can push a "workaround" section on documentation...
see below my email

=======
Reindexing has been tough for me too, so I had to write few wrapper classes and tons of debugging to make sure I didn't miss anything, but what I found out was that, some of my transactions were still opened, remember, titan/Janus open several threads and they might have opened transactions that you don't know about...
I'd suggest the following, if you can:
1) stop all your app servers
2) use gremlin console or write a simple java program
3) execute steps 29.1.5.1, http://docs.janusgraph.org/latest/index-admin.html
3.1) set this storage.parallel-backend-ops=false
3.2) set this ids.num-partitions=1
3.3) depending how large your graph is, this shouldn't take long...
4) once your indexes are done, bring up your servers.

That should do the trick, it did for me...
Enjoy,
Rafa