That's because, in the case of:

perl -e 'print "a line\n" x 10' | less +F
[download]

the end-of-file is encountered as soon as the program finishes, so the pipe gets the end-of-file, and less knows it has received all the input it's going to get.

Here's a good way to test it -- try executing the following command:

perl -e 'while (1) {print "a line\n" x 1000; sleep 1}' | less
[download]

Now the input buffer to less will fill up almost immediately, so you'll see the output right away.  However, if you execute the command "G", it will never find the end-of-file, because the end-of-file hasn't been written yet!

Similarly if you execute the same command with less +G at the end instead of less, you'll never see output for the same reason.