I am controlling them in the same way I would control them if they where individually connected to a shift register.

Having one long shift register or an n-times shorter with multiplexing shouldn't make much of a difference.

Multiplexing means the job has to be done for each scanline instead of for each frame, but each scanline contains less information so the total amount of work to do should be about the same.

That's why I asked if I had missed something important.