How to model my rooftop vertical on 4NEC2?

Question

A few months ago I built a coil loaded vertical, and put it on my rooftop. My roof is made of corrugated metal, so I connected GND to it, to act as a ground plane (or is it a counterpoise?). With some shunt matching I got a decent antenna.

Out of curiosity, I tried to simulate it on 4NEC2. I started with one of the example files included wih it, the GndScreen.NEC file. This file has 16 radials spreading out. I assume "a ridiculous amount of radials" would be the same as a solid ground plane, at least for HF. So I sized the radials to the shorter side of my roof, changed the radiator to match my radiator length, copied the radials and shortened, to emulate a capacitive hat, and added a load 3/4 of the way up, like I have with my vertical, and elevated it 6M off the ground, like my roof is. The end result looks like this:

The frequency plot matched my empirical tests very closely. But when I tried running the Far field analysis for 3.5MHz, I got this:

This antenna is showing only 30% radiation efficiency. What causes it to be so low? The radiator is about 8M long and it has a 30uH coil 3/4 of the way up. The radiator for a full-sized quarter wave monopole would need to be 20M tall. Mine is 8M. Is a radiation efficiency of only 30% the expected result for a radiator slightly less than half its ideal size?

Here's my NEC file:

CM 
CE
SY len=8.8  'Total wire length
SY hgh=9    'Tower heigth
SY segV=int(hgh)*5  'Vertical nr of segments
SY segH=int(len-hgh)    'Horizontal nr of segments
SY ra=360/16, radl=10   'Nr radials, radial length
SY radh =6  'radial height sabove ground
SY rseg = int(radl) '(integer) number of segments
SY cu=5.8e7, fe=1.39e6  'Wire loading for Copper, Steel
SY wrad =.0015  '1.5 mm2 house-wiring
SY RL=0.5
SY lenf=7.5
GW  1   10  0   0   radh    0   0   lenf+radh   .015
GW  11  rseg    0   0   radh    radl*cos(1*ra)  radl*sin(1*ra)  radh-0.6    wrad    'Ground screen
GW  12  rseg    0   0   radh    radl*cos(2*ra)  radl*sin(2*ra)  radh-0.6    wrad
GW  13  rseg    0   0   radh    radl*cos(3*ra)  radl*sin(3*ra)  radh-0.6    wrad
GW  14  rseg    0   0   radh    radl*cos(4*ra)  radl*sin(4*ra)  radh-0.6    wrad
GW  15  rseg    0   0   radh    radl*cos(5*ra)  radl*sin(5*ra)  radh-0.6    wrad
GW  16  rseg    0   0   radh    radl*cos(6*ra)  radl*sin(6*ra)  radh-0.6    wrad
GW  17  rseg    0   0   radh    radl*cos(7*ra)  radl*sin(7*ra)  radh-0.6    wrad
GW  18  rseg    0   0   radh    radl*cos(8*ra)  radl*sin(8*ra)  radh-0.6    wrad
GW  19  rseg    0   0   radh    radl*cos(9*ra)  radl*sin(9*ra)  radh-0.6    wrad
GW  20  rseg    0   0   radh    radl*cos(10*ra) radl*sin(10*ra) radh-0.6    wrad
GW  21  rseg    0   0   radh    radl*cos(11*ra) radl*sin(11*ra) radh-0.6    wrad
GW  22  rseg    0   0   radh    radl*cos(12*ra) radl*sin(12*ra) radh-0.6    wrad
GW  23  rseg    0   0   radh    radl*cos(13*ra) radl*sin(13*ra) radh-0.6    wrad
GW  24  rseg    0   0   radh    radl*cos(14*ra) radl*sin(14*ra) radh-0.6    wrad
GW  25  rseg    0   0   radh    radl*cos(15*ra) radl*sin(15*ra) radh-0.6    wrad
GW  26  rseg    0   0   radh    radl*cos(16*ra) radl*sin(16*ra) radh-0.6    wrad
GW  11  3   0   0   lenf+radh   RL*cos(1*ra)    RL*sin(1*ra)    lenf+radh   wrad    'Ground screen
GW  12  3   0   0   lenf+radh   RL*cos(2*ra)    RL*sin(2*ra)    lenf+radh   wrad
GW  13  3   0   0   lenf+radh   RL*cos(3*ra)    RL*sin(3*ra)    lenf+radh   wrad
GW  14  3   0   0   lenf+radh   RL*cos(4*ra)    RL*sin(4*ra)    lenf+radh   wrad
GW  15  3   0   0   lenf+radh   RL*cos(5*ra)    RL*sin(5*ra)    lenf+radh   wrad
GW  16  3   0   0   lenf+radh   RL*cos(6*ra)    RL*sin(6*ra)    lenf+radh   wrad
GW  17  3   0   0   lenf+radh   RL*cos(7*ra)    RL*sin(7*ra)    lenf+radh   wrad
GW  18  3   0   0   lenf+radh   RL*cos(8*ra)    RL*sin(8*ra)    lenf+radh   wrad
GW  19  3   0   0   lenf+radh   RL*cos(9*ra)    RL*sin(9*ra)    lenf+radh   wrad
GW  20  3   0   0   lenf+radh   RL*cos(10*ra)   RL*sin(10*ra)   lenf+radh   wrad
GW  21  3   0   0   lenf+radh   RL*cos(11*ra)   RL*sin(11*ra)   lenf+radh   wrad
GW  22  3   0   0   lenf+radh   RL*cos(12*ra)   RL*sin(12*ra)   lenf+radh   wrad
GW  23  3   0   0   lenf+radh   RL*cos(13*ra)   RL*sin(13*ra)   lenf+radh   wrad
GW  24  3   0   0   lenf+radh   RL*cos(14*ra)   RL*sin(14*ra)   lenf+radh   wrad
GW  25  3   0   0   lenf+radh   RL*cos(15*ra)   RL*sin(15*ra)   lenf+radh   wrad
GW  26  3   0   0   lenf+radh   RL*cos(16*ra)   RL*sin(16*ra)   lenf+radh   wrad
GE  -1
LD  5   1   0   40  37700000    'Left wire
LD  0   1   6   6   0   30e-6   0
GN  2   0   0   0   13  0.005
EK
EX  0   1   1   0   1   0   0
FR  0   0   0   0   3.5 0
EN

Answer 1

RE: This antenna is showing only 30% radiation efficiency. What causes it to be so low? ... Is a radiation efficiency of only 30% the expected result for a radiator slightly less than half its ideal size?

The value for the Radiat-eff. result in the 4nec2 display includes losses based on the amount of the originally-radiated r-f energy remaining after propagating along an infinite, flat ground plane having the conductivity defined for the wire model as analyzed.

Using 4nec2 and running your unedited .nec file after selecting the "Run Average Gain Test" choice when setting up the NEC engine returns its true radiation efficiency (see below):

Edit of 30 March 2020 w.r.t. the comment of hjf this date:

The antenna defined in your NEC model radiates nearly 100% of the Z-matched power across its feedpoint terminals. After that its radiated fields are reduced by the losses of the propagation environment, but those losses are not attributes of the antenna system, itself.

The following graphic from 4nec2 shows the radiated fields vs. height AGL at a horizontal distance of 50 meters from the antenna defined by your NEC model.

There is significant attenuation in its fields due to the lossy ground plane, even at that short distance. However it is far less than implied by the ~31% value for Radiat-Eff reported by a NEC "Far Field" (only) analysis.