I have to agree with Tom (he got to you first). I've read a large number of people who suggest that the halo effect is the cause for this and that mysterious behaviour, primarily because it's the only effect they know. The halo effect is widely misunderstood, so I'll describe it briefly, though it is almost certainly not your problem.
The halo effect proper primarily effects metals like iron and zinc, which particalise and disperse into the nearby soil. I recently dug up a nail that looked like it had a large ball of iron attached to it. When I scratched away at the ball, it crumbled as highly ferrous soil (obvious silicates mixed in with it). That is the real halo effect. The iron actually leeched into the soil and became dense enough to organise against itself in a crystaline manner. In effect, the nail became greater in volume.
There is also the wet cell effect. It is commonly mistaken for the halo effect, but only applies in the presence of water. When the ground is wet, a conductive (metallic) object will be electrically exposed to the less conductive ground matter, creating an extremely weak wet cell battery. When the energy from your search coil hits this battery, it responds as an electrical system instead of just a conductive object. So, the signal you get is bigger than the signal from the object by itself. That's why everything hits harder when the ground is wet.
There is also the loose soil effect. This is the one that is responsible for the disappearing behaviour and makes newbies' test beds fail. Natural ground is filled with minerals that have been electrically organised by draining water (because the water molecule is dipolar). In some soils, this makes the soil an even better field conductor than the air. When the soil is disturbed, the minerals become disorganised, which causes a scattering effect on the field, and air pockets are introduced into the soil, which causes electromagnetic fields to try to worm their way through the paths the least resistance (the air pockets) instead of conducting through the soil. So, once the soil is disturbed, your potential detection depth can be cut by as much as half. That's why sometimes you can get a deep signal, start digging, it disappears, you keep digging, and it comes back.
As you can see, these soil effects are probably not your trouble. I would highly suggest getting a Garrett Pro Pinpointer whenever time and money permit. If you're pinpointing with a pinpoint feature built into your metal detector, try instead turning your sensitivity all the way down and using the old x-marks-the-spot technique (scan in short, tightly-spaced swings in one direction, then turn 90 degrees and scan in the other direction; keep your eyes on the ground, not the coil; mark the spot with your eyes when you're confident).
