I don't want worry you but March-Dollase is only working for axial crystal system since you need to define ONE preferred direction in order to replace the plane multiplicity (multiplicity factor by an "effective multiplicity" for each plane. This means that for cubic it does not work anymore. PowderCell certainly calculates something but the use model is inappropriate. 

To consider an assumed texture or use a measured one is also quite speculative since the texture of the current state of ferrite cannot be used since it is for another phase. Moreover, the reason, why these retained austenites are still existing and all other are gone has perhaps also something to do with the orientation or misorientation of these grains. This means that also a high-temperature texture cannot be simply used. 

On the other hand....if you analyse textured material the difference in phase fraction is not that remarkably difference as long as you use not only 2 or 4 peaks. The reason is that intensity usually does not "disappear", i.e. intensity of 002 with be reduced (since the respective grain does not reflect intensity into the detector since it is rotate by texture). But the same grain is now in reflection condition for another plane and contributes to another interference, e.g. h0l. 

Moreover, to concider texture for phase mixtures requires a certain amount. The concentration of retained austenite is from my point of view to low and statistically not significant. To iimprove this you need much longer measurements...and more reflections.

If your steel is magnetic then retained austenite could be accurately measured using magnetic measurement.

If you want to stick with XRD then use March-Dollase approximation to take care of texture effect using anysuitable software (e.g. Powder Cell). This will definitely improve your phase quantification result.

Well, you can just do some proper diffraction measurements of texture. If your sample is pretty large then neutron diffraction could help you. Since I am from JINR, Dubna, I would advertise SKAT texture diffractometer (e.g., doi:10.1107/S1600576714015830). It is accessible to users via the standard system of proposals. Normally, a few hundreds of time-of-flight neutron diffraction spectra will be measured from the sample (you will surely see at least 5 peaks for ferrite, maybe more), allowing rather precise simultaneous phase and texture analysis. If the sample is small then use X-rays (with cradle) or synchrotron diffraction...

Thank you for your answers! I will look into the March-Dollase approximation. I assume I will need to know or simulate the texture of the austenite phase to benefit from this method.

Since the diffracted intensity depends on the texture and phase fractions both you need to measure texture using Neutron or even XRD with goniometer and then apply March-Dollase approach. MAUD (free software) also offers the March approach. f(g) of texture function is required.

I don't want worry you but March-Dollase is only working for axial crystal system since you need to define ONE preferred direction in order to replace the plane multiplicity (multiplicity factor by an "effective multiplicity" for each plane. This means that for cubic it does not work anymore. PowderCell certainly calculates something but the use model is inappropriate. 

To consider an assumed texture or use a measured one is also quite speculative since the texture of the current state of ferrite cannot be used since it is for another phase. Moreover, the reason, why these retained austenites are still existing and all other are gone has perhaps also something to do with the orientation or misorientation of these grains. This means that also a high-temperature texture cannot be simply used. 

On the other hand....if you analyse textured material the difference in phase fraction is not that remarkably difference as long as you use not only 2 or 4 peaks. The reason is that intensity usually does not "disappear", i.e. intensity of 002 with be reduced (since the respective grain does not reflect intensity into the detector since it is rotate by texture). But the same grain is now in reflection condition for another plane and contributes to another interference, e.g. h0l. 

Moreover, to concider texture for phase mixtures requires a certain amount. The concentration of retained austenite is from my point of view to low and statistically not significant. To iimprove this you need much longer measurements...and more reflections.

Just to add a few to above comments;

U cant use only single data to get texture information as it anisotropic. To getbetter texture information u need to rotate and tilt ur sample if u had compact cradle sample stage, but if not than rotate the sample by certain fix angle and do scan for each step.

MAUD is good free software to handle texture (M/D, SH, WIMV, EWIMV, etc) . U can get ODF, Pole Figure and Inverse Pole figure.

Hope it help