Figure 5 from Martin Simons shows pretty much what I was talking about. In that figure, the tail load is zero, and as as stated by Ben and Paul above, that can only be for one specific flight condition. Since the Cm about the wings AC is constant, an increase in angle of attack is going to cause only the wing lift coefficient to increase, which in turn produces a nose-up pitching moment about the CG higher than the wing's Cm, so that the tail will then have to lift up for trim. At a lower AoA, the situation is reversed: Cm about the wings AC is still the same, but now lift is reduced and there is a net nose-down pitching moment about the CG, requiring a downward load on the tail.

This is the case I tried to explain in my first post, namely that it can happen that the tail loads down at higher speeds (low AoA) and up at lower speeds (high AoA). It seems counter-intuitive since you need up elevator at low speeds, but you tend to forget that the whole horizontal tail is at an angle of attack, so that even with up elevator the tail may still be lifting up. I think that is what Ben was also trying to explain in his earlier posts. It is also possible to design it with the load always up or always down inside the airplanes flight envelope.

The possibility of an upload on the tail for trimmed flight, or zero load as in Figure 5, is also dependent on placing the CG behind the wings AC as shown in the figure. (This condition of course assumes that the wing has a negative pitching moment about its AC). If the CG is in front of the wings AC, the tail will indeed always have to lift down for trimmed flight, since the pitching moment due to the lift and the wing pitching moment will both be nose-down. In all of this, the CG also have to be ahead of the airplanes neutral point for positive stability, but the NP can be far enough behind the wings AC that you can still have the CG itself far enough behind the wings AC to trim with an upload on the tail.