I totally agree.
My version is based on yours with millis(). I think, if you're going to use millis(), than use it properly. If your version had worked with delay() from the beginning, my millis() version would probably not have been created here at all.
But we had clarified this millis() <-> delay() problem in the meantime 