Tubes can add some sweet sounding harmonics to recording that need them. They can also add these harmonics to recordings that do not need them. It used to be said that James Parker of Audio Research specifically added 2nd order harmonics to sweeten his design. As a scientist of sorts, I prefer to add harmonics on an as needed basis.This sort of blanket statement is problematic! The actual psychoacoustic situation is much more complex. Just as tubes 'add harmonics', so do transistors (although to a lesser extent). The problem is that with the vast majority of transistor amps (greater than 99%, due to insufficient gain bandwidth product, which is a discussion for another thread) the harmonics they add are higher orders which are interpreted by the ear as brightness and harshness.
The ear has a masking principle, and in addition is relatively insensitive to the lower ordered harmonics (2nd-4th). If the lower orders are present in sufficient quantity, they can mask the presence of the higher orders. This is why tubes sound smoother than solid state despite having higher distortion.
On the personal preference side, I enjoy the greater soundstage width and depth, instrument timbres and low level detail that IME that is commonly associated with tube circuits.