Two issues come to mind.
1) Antenna elements are engineered to a specific length and diameter so that their impedance will allow efficient energy transfer. Placing other conductors close to the antenna changes the impedance of the antenna elements. Metal roofing, duct work, mast, cable, another antenna, etc., will all have greater effect when they are placed closer to the antenna. A general 'rule of thumb' may be 'try to maintain at least one wave length separation and avoid situations where obstructions are in front of the antenna.'
Wavelength (Meters) = 299.792458 / Frequency (MHz)
2) A metal roof will reflect radio waves more efficiently than wood or composition shingles. The angle of the incident wave and the shape of the reflective surface will determine the direction the reflected wave travels away from the reflector. Flat surfaces tend to reflect coherently (the reflected light, sound or radio wave tends to stay together rather than scatter and dissipate) like a mirror. If a signal reflects off the roof and then strikes the antenna, some amount of multipath interference is produced. When the reflector is very close it's doubtful that the delay would be significant enough to cause trouble for the tuner/decoder. However, if the reflected signal arrives out of phase, it will tend to cancel the direct path signal. It's also possible for the signal to arrive in phase and add to the signal. If there's a significant amount of 'reflector action' taking place, you would see significant changes in signal strength as you move the antenna up or down (though the distance of a wavelength).
The bottom line is, the real world answer to "Are you reliably receiving the signal you want?" is more important than the theory world answer to "What's a good rule of thumb?"
If the well is dry and you don't see rain on the horizon, you'll need to dig the hole deeper. (A fancy pump won't make a dry hole wet.)