I don't know to what extend the ions in your simulation aggregate or at which temperature your simulations are performed, so I cannot tell you with certainty if what you observe is in fact correct or an artifact. 

However, I'm pretty sure that there are papers that look at concentrated (or even brine) solutions. So it should be easy enough for you to confirm if your behavior is realistic.

If it is not consistent with the literature, it might have to do with the forcefield or a mistake in the input scripts.

If it is consistent with the literature, I can offer the following insight: The fact that K and Cl are reasonably comparable in size makes that they tend to form more ion pairs than for example LiBr. This is a consequence of the balance between ion-ion and ion-water (or perhaps ethanol) interaction strengths. The large ion concentration also results in more pairing between the ions.

http://scitation.aip.org/content/aip/journal/jcp/141/12/10.1063/1.4896380

Article Structure and transport of aqueous electrolytes: From simple...

Thanks very much, Dr Hartkamp. My simulations were performed at 300 K,1 bar.

1.5M is a lot of ions. I experienced ion aggregation with ffamber99sb at physiological conditions. Taking a different set of ions fixed it. Joung,J.Phys.Chem.B,Vol.112,No.30,2008

Unphysical ion aggregation is indeed a known problem of some forcefields. Forcefield parameters for ions are typically optimized based on the ion-water interaction, where the water is a modeled with a particular model. Targets for the optimization can for example include the radius the hydration shell, or the solvation free energy. These 'optimal' parameters are not necessarily representative of realistic aggregation behavior. The "Joung and Cheatham" paper is indeed one of the nice ones to consider for monovalent aqueous solutions. I'm not sure about ethanol solutions though.