Hello,
When setting the frequency range from 1 to 10 GHz and the number of frequencies that are used for simulation is too small, the interpolator algorithm for plotting the impedance on the smith chart may go wrong. Even intelligent adaptive algorithms may fail.
You need sufficient frequency simulation points around your resonant frequency. Check the actual frequencies that are used for the simulation.
It is correct that you get several resonances. The lowest resonant frequency is where the electrical length is 0.5 lambda (that means that the physical length is somewhat less depending on the thickness of your dipole). Impedance will be between around 50…70 Ohms. Thicker dipoles (that have size remarkably below 0.5 lambda are closer to 50 Ohms).
The second one appears around 1 lambda, that impedance is high end depends strongly on thickness/wavelength ratio. Very thin full wave dipoles will be around kOHms, where very thick ones (for example thickness/length = 0.2), will have several hundreds Ohms.
You should not see a resonance at 3.3 GHz. I think you should use more frequency points that are actually used for simulation and/or check the meshing.
Regarding meshing, there must be something that will show you how the simulator divided your structure before simulation. In IE3D it is the "show meshing" command. Maximum mesh size of 1/20 or 1/10 lambda is mostly used as default. Correct meshing is vital to get best results with minimum simulation time. Where you expect strong current amplitude change versus position, you need a finer mesh. You can be lucky, with your dipole, there are no strong current changes versus position, so a mesh size of 1/10 to1/20 lambda will give you practically useful results.
I hope someone else will comment on how to find/set meshing parameters (and other things) as I don't know HFSS.