I think I found this bug while runningg the following script in gnuplot 5.0 patchlevel rc2. The script I tried to run is,
set samples 1000
F0 = 1.0
K = 9.0
wn = 3.0
H(w,zt) = (F0/K)/sqrt((1-w/wn)**2+(2*zt*w*wn)**2)
Phi(w,zt) = atan(2*zt*w*wn/(wn**2-w**2))*180/pi
set grid
set multiplot layout 1,2 title 'Frequency Response' columnsfirst
set title 'Magnitude'
set xlabel 'w (rad s^{-1})'
set ylabel 'H (m)'
set xrange [0:10]
plot H(x,0.005) t '{/Symbol z}=0.005',H(x,0.01) t '{/Symbol z}=0.01',H(x,0.2) t '{/Symbol z}=0.2',H(x,0.5) t '{/Symbol z}=0.5',H(x,1.0) t '{/Symbol z}=1'
#
set title 'Phase'
set xlabel 'w (rad s^{-1})'
set ylabel '{/Symbol p} (deg)'
set xrange [0:10]
plot Phi(x,0.005) t '{/Symbol z}=0.005',Phi(x,0.01) t '{/Symbol z}',Phi(x,0.2) t '{/Symbol z}=0.2',Phi(x,0.5) t '{/Symbol z}=0.5',Phi(x,1.0) t '{/S ymbol z}=1'
unset multiplot
In gnuplot 5, I found that one function is plotted in both the subplots and the other, in only one. I have attached the qt output.
However, this issue did not arise when I tried plotting it with gnuplot version 4.6 patchlevel 6 (running in x11 mode). The issue did not arise even with the x11 terminal in gnuplot 5 itself.
Version 5.0 rc2 was a beta-test version from Aug 2014 that was replaced a few weeks later by -rc3 and then the by first general release of 5.0 in Jan 2015.
I have not tried to reproduce your test script on the -rc2 beta version, but I get correct output using the 5.0 base release, using 5.0 patchlevel 1, or using the current development version 5.1. So although I do not see any obviously relevant entry in file, this issue may be something that was fixed during the beta test period.