This is likely to be a inadvertent simulation mistake. If you simulate over (say) 0.5GHz thru 4GHz, you should see the region where the antenna works in the middle, and the regions where the device is no longer resonant.
if you used a separate instance to simulate the second (1-3 GHz) check that you got the units correct. Maybe you inadvertently had it set to MHz instead of GHz.. or you had a boundary symmetry different, or some little thing like that.
- - - Updated - - -
I was able to check with a clever colleague. There is no mistake. Both frequency ranges give the same result.
Mismatch is about S11= -0.5dB from 1.5Ghz to 2.7GHz. The resonance is at about 3.14 GHz where S11 is about -5dB (poor).
The 3Ghz curve follows the 3.5Ghz curve, but of course, the resonance is beyond its range.