A source is spread out over an entire segment. So when you change the
number of segments, you change both the length and the effective
position of the source. When the source is at the bottom of a quarter
wavelength radiator, small changes in source position don't make much
difference in the impedance seen by the source. However, when the
antenna approaches a half wavelength, the source impedance changes quite
dramatically with source position. Consequently, you'll see substantial
changes in reported source impedance with segmentation in that case.
This might or might not be the cause of what you're seeing. As an
experiment, you might try moving the source up one segment and see how
big a difference it makes.

Whenever the result is very sensitive to small changes in the model, you
shouldn't expect a real antenna to come out exactly like the model
predicts, since small differences between the model and real antenna
will likewise cause significant differences.

The absolute length doesn't matter -- a 146 MHz antenna will be no more
or less sensitive to the same amount of change (in terms of percentage
of the antenna size or of the wavelength) than a 3.5 MHz antenna if both
are proportioned the same. In fact, 146 MHz antennas are typically
considerably fatter in terms of wavelength than 3.5 MHz antennas, and
this makes them less sensitive to small changes.

Roy Lewallen, W7EL

wrote:
I found that increasing the number of segments had a significant change
in the input Z. The material I read on 5/8 antennas indicated the real
part of the Z was near 50 ohms. I could not get that result until I
increased the number of segments. Guess it is cause 146 MHZ antennas
are a good bit shorter than 3.5 MHZ antennas, and any small deviation
such as lenght, or # of segments will change the end results.
Gary N4AST