How does the createFeed() by GUI (antenna toolbox and selecting a closed polygon) work as an excitation source?

I am experimenting with MATLAB antenna tool box. The simulation result of a geometry differs quite a bit when I use a simple delta-gap source feed model(one edge) vs. select a closed polygon from the GUI. I am looking for a rather elaborate explanation on 1)how does the select a closed polygon from the GUI feeding method work? I don't get to see a lot of documentations on it. 2) if i use the command line objectname.createFeed(location), how does it define the feed edges? 3)which one should I use in the case for a 3D custom antenna input as .stl file to reflect an accurate output?

Here is the breakdown of my experiment:

Baseline:

I use a dipoleMeander antenna from the antenna toolbox.

dm = dipoleMeander
sobj = sparameters(dm,sim_freq_band, Z0);

I directly run an S-paramter analysis and plot the result. This result is my baselines to compare, it is the simple delta-gap source feed as described from the documentation.:

Here is the plotted antenna, when zoon in, you can see at the middle of the feed dot there is one edge defined which is the edge used for the delta-gap model.

s-parameter value:

This is the baseline result I will compare with the selecting a closed polygon feed approach.

Export to .stl file and import it back. Create the feed using GUI:

I repeat the same step, this time instead of using one edge, I open the GUI and select a closed polygon:

dmcd = customAntennaStl('Filename', 'meaner_dipole.stl', 'Units', 'm','UseFileAsMesh',0); 
dmcd.createFeed();

The here is my edge selection, I chose a quadrilateral around the point [0,0,0]:

I then obtain the s-paramter result as well as my current distribution result:

Now I am seeing quite a different s-parameter value. Both the optimal frequency and the magnitude shift quite a lot.

Sanity check. The .stl model with one edge feed:

As a sanity check, the result using the imported .stl file returns the same reult as the baseline method when using a one edge delta-gap feeding model, suggesting that the difference in s-parameter results is likely due to the feeding style only.

mesh(dm);
stlwrite(dm,'meaner_dipole.stl');
dmc = customAntennaStl('Filename', 'meaner_dipole.stl', 'Units', 'm','UseFileAsMesh',0);
dmc.createFeed([0.00,0,0.00],1);
sobjc = sparameters(dmc,sim_freq_band, Z0);

Only one edge in the middle of the feed dot for the exported .stl model, and it yields the same result as the baseline:

Extending the problem to a custom 3D metal:

The reason for the above investigation is that I would need to excite 3D custom antennas (imported through .stl files) and yield results as realistic as possible. Here is an example of my problem:

Let's say I create 3D curved metal, and I excite it at the very center ([0,0,0] is the center in the following example):

% Step 5: Load and show the shifted STL file using customAntennaStl
c_shifted = customAntennaStl('Filename', 'test_conform_shifted.stl', 'Units', 'm', 'UseFileAsMesh', 0);
c_shifted.createFeed([0.00,0,0.0],1);
mesh(c_shifted);
current(c_shifted,900e3);

In this case the center of the feed dot is not even at an edge, it is centered at a node(vertex), here is what it looks like when zoon in:

Another way is to excite it using GUI at the very center (a closed polygon around the point [0,0,0]).

c_shifted = customAntennaStl('Filename', 'test_conform_shifted.stl', 'Units', 'm', 'UseFileAsMesh', 0);
c_shifted.createFeed();
current(c_shifted,900e3)

I ran a current distribution analysis for both approaches, the results are drastically different:

When the input feed dot is at the node(vertex):

When input feed is selected as a closed polygon:

Anyone knows how does that select-a-close-polygon feeding method work? Is it a derivative of the delta-gap source feed model as shown in the first example? MATLAB seem to not support coax feeding method for customAntennaStl model so I guess I can rule that one out?

Cheers