Could you clarify your question a bit:
1. Are you seeing more randomness than you would expect in your initial state distribution?
2. Less randomness than you would expect?
Unless you select the checkbox to “Calculate from distribution” checkbox, the model interprets your inputs for initial conditions as a frequency distribution so you should see some randomness in your initial conditions. The more cells that you run (i.e., the larger the population you are sampling from), the closer that your simulated initial conditions will match your input values.
If you choose to “Calculate from distribution” for the number of cells, your initial conditions will be deterministic allocating the value in the “Relative Amount” field to be the number of cells corresponding to each record. You can find out more in the documentation for this property.