Here's another way with sed that replaces only the first empty line without using \n in the RHS or the gnu sed 0 address extension:

sed '/^$/{      # if line is empty
x               # exchange pattern space w. hold space
//{             # if pattern space is empty (means the hold buffer was empty)
s//%/           # replace it with a % character
h               # overwrite hold space (so now hold space looks like this: ^%$)
s/$/ ghi/       # add a space and the string ghi after the %, then
G               # append content of hold buffer to pattern space so now the
}               # pattern space looks like this: ^% ghi\n%$
//!x            # if pattern space is not empty it means a change was
}               # already made so exchange back (do nothing)
' infile

one liner:

sed -e'/^$/{x;//{s//%/;h;s/$/ ghi/;G' -e'}' -e'//!x' -e'}' infile

Here's another way with sed that replaces only the first empty line without using \n in the RHS or the gnu sed 0 address extension:

sed '/^$/{      # if line is empty
x               # exchange pattern space w. hold space
//{             # if pattern space is empty (means the hold buffer was empty)
s//%/           # replace it with a % character
h               # overwrite hold space (so now hold space looks like this: ^%$)
s/$/ ghi/       # add a space and the string ghi after the %, then
G               # append content of hold buffer to pattern space so now the
}               # pattern space looks like this: ^% ghi\n%$
//!x            # if pattern space is not empty it means a change was
}               # already made so exchange back (do nothing)
' infile

one liner:

sed -e'/^$/{x;//{s//%/;h;s/$/ ghi/;G' -e'}' -e'//!x' -e'}' infile

Though really, this is piece of cake for ed:

ed -s infile <<< $'/^$/s//% ghi\\\n%/\n,p\nq'

replace ,p with w to edit the file in-place.