Hi Capri,
as you are correctly stating, simulation of charge injection only gives a qualitative feel of what's going on. As the models are not initially derived for transistors operating as a switch, the results are not totally accurate (at least AFAIK).
The proper way to handle this problem is to design a circuit with signal independent charge injection (which leads to an offset only). Even better, signal independent charge injection in a fully differential circuit is 1st order cancelled.
For more information on this solution, have a look at Gregorian-Temes
Analog MOS Integrated Circuits for Signal Processing
by Roubik Gregorian, Gabor C. Temes
and Johns-Martin:
http://www.designers-guide.org/Books/#Johns-1996Paul