Higher is always better, but I would dispute that the car body somehow
"shields" an antenna (unless the antenna is butt up against the cab,
say). More often than not, the favored direction (the major lobe) is
found aligned in the direction of the most metal. That is, with a
bumper mount or fender mount, the best propagation is corner to corner
across the car NOT away from the car. In other words, the sheet metal
of the body supports and favors lobe development (instead of the metal
acting as reflector). I've noted this documented by QST since the
'60s.

73's
Richard Clark, KB7QHC

I won't get into the method but it has a lot to do with the frequency and if
the antenna is a full 1/4 wave, 5/8 , or longer co-linear antenna as to what
the patern is. The patern can be over the longest metal path for a 10 meter
whip or it can be "blocked' and go in the opposit direction for an antenna
mounted on one side of the trunk of a car for a co-linear 440 mhz antenna.
This is from the antenna makers diagrams and polar plots. I have observed
by leaving a car in one spot and changing from a 1/4 wave, 5/8 wave and a 6
foot long colinear 2 meter antennas on the same mount that each can have a
diffrant patern , especially where working repeaters and the repeater
antennas are at differant heights. For the areas I am in , I would rather
have a 1/4 wave in the center of the roof , but have settled for a 38 inch
long colinear type dual band antenna on one side of the trunk of my car and
not worried about it.