Ok, I admit having to put the l on the chlorine makes the headline even more obtuse than Jacks, but I have seen split peaks too when faced with sample extracts that contain a 2-solvent system. This was analyzing semivolatile soil extracts which had not been reduced in volume low enough to get a solvent exchange into dichloromethane, and thus a dual solvent extract was applied to the chromatograph. These samples had variable percentages of acetone and dichloromethane, and the early-eluters split into peaks of relative areas based on the solvent composition and the soluability of each analyte into these two solvents.
I also witnessed this phenomenon when performing direct aqueous injections on low-boiling compounds, which may be more similar to what you are seeing. Water does not wet siloxane phases well, and when trying to focus it by applying the 20°C lower rule of thumb, the early eluters had very bad peak shape. This was improved by going to higher starting temperatures.