NEC engine doesn't support any geometry except point-to-poing 2D segments, which have zero diameter (for group delay tracing it acts as zero diameter wire), but it's impedance is later adjusted to actual diameter. This adjustment works only for thin wires, thats why NEC engines are not very helpful at microwaves.
4NEC2 front-end support definining surface geometry as single SP card (instead of GW card for wires), which defines rectangle at once. But NEC simulator suppports only wires, so 4NEC2 convert SP card to dozens of GW card befrore executing NEC. NEC solves model using wire mesh for SP surfaces.
It's the same as "Run - Geometry builder - Patch" and generate GW cards
If you want to make antenna structures, other than reflector, from strips (any non-cylinder geometry) - your capabilities are very limited. For some open-end elements like Uda-Yagi elements - it can work, if you find equivalent diameter. DL6WU is videly venerated for his conversions of strip-to-cylinder for Uda-Yagi elements.
For non-open-end elements it doesn't work. If you draw loop antenna (e.g. biquad) from strips, NEC2 count signal delay as it traced exaactly at the wire center (applying impedance correction AFTER)
For wide strip (at microwave frequencies everything is wide) it gives big error, as electrical length (delay) is not obvious, as it contain distributed capacitance/inductance
[biquad, 1/4 is shown]
I do simulations using 3D software. For some antenna types (especially Uda-Yagi) great job was done using NEC-based engines. To convert wire-type antenna to strip-type (including PCB) I find transformation ratio.
I draw wire (cylinder) L/2 dipole in 3D and calculate resonant (jX = 0) frequency.
Than I draw strip L/2 dipole of desired shape and ajdust L to make it resonant again.
k=L1/L2, multiply original element widths by K.
This method is not good for metal boom, as it intersect with split L/2 dipole.
Specifically for Uda-Yagis I calculat k directly in 3D simulator.
I draw geometry with k variable multiplied to each particular element.
Then I sweep k with smal step (1% or less) and look at Directivity plots. I take k from best plot.
This is extremely easy approach and work with ANY elements and with ANY boom (boom-correction and element shape correction work for virtually any combination of size, shape and mounting method).
Calculation may take few hours, but it takes only couple minutes of my time to setup simulation
NEC engine work good with open-end elements (like dipoles. Uda-Yagi), work fairly good with loops from very thin wire, work bad with loops from thick wire (any wire > 1 GHz) and doesn't work at all with transmission lines. Zo impedance of such line is a fixed ratio from D/d (diameter to distance), but NEC cannot calculate it, as it treat d as zero diameter.
Thats why you'll never see NEC models with transmission lines, or they will have very high impedance >400 Ohm, as have very thin wires spaced to large width.