You could try touchdown PCR (TD-PCR), where you start with a high melting temperature and then after each cycle the melting temperature drops by one degree. This reduces the chance of off-target amplification.

But your primers are very long and you're already using a high melting temperature. Do your primers contain overhangs that you're adding so the PCR product contains overlaps for Gibson assembly? If so, you shouldn't include the overhangs when calculating Tm, since in the early cycles those overhangs won't bind to the plasmid template.

Are you sure there is no chance the primers are binding to a second location in the plasmid? You should check if the 3' end of your primers can bind elsewhere. Small amplicons amplify much more efficiently than long amplicons in PCR, so even a small chance of off-target binding by your primers will mean the 500bp amplicon gets amplified instead of the 10kb amplicon.