Depending on the frequency in use, you might be
doing pretty good to be able to measure variations
within 1.5 db anyway.
There are lots of variables
here. I take it your source dish is transmitting
a signal generator output horizontally, and you
switch between the horn and your antenna under test
at the other end of the range (is that near, or
some kilometres away)?
The gain lobe in the
source may be high enough that even a slight wind
pressure would be enough. Depending on frequency,
rain or air moisture could affect it. If your antenna
under test is also a dish, you have a double chance
of wobbling the lobes 1.5dB worth. Everything has
to be fixed firmly, and not be prone to vibration.
I was once careless with a piece of heliax coax
cable, and caused more than 1.5dB of problems because
of bending it too hard!
All of the above
assumes you are using very stable calibrated kit.