The problem with the argument that roguelikes must be original to attract attention is that people often reject completely original roguelikes. In fact, most of the popular roguelikes are fantasy themed. And in general fantasy themed roguelikes do much better than more original themes. There have been lots of experimental roguelikes over the years, and practically none of them survived.
I myself developed a fairly original roguelike that was based on the dreams, but no one cared. There's also Privateer Ascii Sector, Triangle Wizard, Prospector, Gearhead 2, Unrealworld and MegamanRL to list a few. All of which used highly original ideas, yet did fairly poorly in terms of attracting an audience and starting discussion. Also, keep in mind that Serial Killer RL's ideas were not original. There are many TV shows, movies, books and comics based around serial killers. JTHM and Dexter are two very popular ones, American Psycho is another.
The only reason why this project received as much feedback as it did was because of how shocking it was. It was the shock value, along with the "decent" graphics that caused all of this discussion. Even the other fake roguelike, MnemonicRL didn't attract as much attention because it wasn't as shocking.
Honestly, I don't know what the secret to attracting decent attention to your roguelike is. It seems graphics play a large part, as does how shocking the roguelike is, but beyond that I have no idea. Even many roguelikes with good graphics have failed to attract much attention.
I would also like to suggest that perhaps how complete and complicated a game looks plays a factor. Serial Killer RL looked like a fairly complex game with a lot of features that looked very close to complete. Perhaps this had an impact, too?
