That's because Ives convinced me to get rid of the first two bytes of peoples' names, which were language identifiers, and let people have 10 character names.
Previously, names looked like this in the data: 0x09 0x45 then 9 characters then a null termination for a total of 12 Unicode bytes.
Now the names look like 10 characters and a null termination. (Language identifier stripped.)
The client is hardcoded by SEGA to write the logs in a way that skips those two Unicode bytes for language identification and writes the rest of the name string after that point.
We can look at hex editing the client to write the log properly, but that's extremely low priority... Unless you want me to change back to 9 character names again with a preceding language identifier.