Is there no way to quantify the losses, no rule of thumb, should we
all
throw our tuners away because they may have unpredictable horrendous
losses?

In general I do....I don't use a tuner for any of my everyday
antennas...

What steps should be taken by the average amateur to obviate these
losses.

Quite simple really. Use the very minimum inductance needed to get
a usable match. I once did tests to determine the appx amount of
tuner loss when tuning properly vs improperly. When using an excess
of inductance to match, I saw tuner losses approach 20%. IE: 20w out
of 100w. When carefully tuning, using the very least inductance, this
loss could be reduced to a value nearly unreadable. BUT!!! This was
on a resonant 50 ohm system. You could expect the losses of an extreme
mismatch to be higher, even if you tuned very carefully. The losses
and
coil warming when feeding a 1/2 size dipole with a T net tuner are well
known.
Thats how I got a free MFJ 989c when a friend of mine couldn't keep
from
frying the plastic coil forms. He bought a nye viking, and gave me the
989c.
Which I've used since with no problems at all....It wasn't really the
tuners
fault...He was abusing it trying to feed a 1/2 size dipole, and then
running a
TL-922 flat out on top of that... MK