How can I get a linear calibration curve for CO2 by GC-TCD?

Several calibration points (from 1% (v/v) to 30% (v/v)) of CO2 in the air are measured, but the response of the detector is non-linear.
I'm using a Focus GC (Thermo Scientific Instrument) equipped with a Hayesep Q column (3 m); the temperatures of the thermal conductivity detector, injector and analytic column are 200°C, 50°C and 50°C, respectively. Nitrogen is used as the carrier gas at a flow rate of 20 mL/min. The eluition is set up in a isocratic mode (about 2.5 min at 50°C because I have to analyze also H2) and then there is a gradient from 50°C to 180°C: after the starting of the gradient within 2 min I have CO2 peak.
Using these parameters, I have a linear calibration curve for H2, but not for CO2. Maybe someone can help me?