When the IR light is bouncing on a materials" surface, a part of it reflects, a part of it is absorbed and a part is transmitted. In FTIR spectroscopy we have the possibility to collect the reflected light or the transmission light. There is no possibility to collect the absorbed light, but we can assume (...the instrument is doing it for us) that if we applied a specific amount of energy and we collected an other amount, the rest is the absorbed light.
The FTIR detector (in standard mode/configuration) measures the amount of light transmitted by the sample. This means that the signal you measure, which is just the ratio of light intensity between the input and output beams, is a combination of the material absorbance and the light reflected from the surface. This reflection loss can be calculated via the Fresnel equations, and should be corrected in the spectrum to give a true measure of absorption coefficient.
I agree with Dr. Musgrave. So, in DRIFT (Reflectance Diffuse), you measure a part of diffuse reflected light over your solid sample, besides, in ATR mode (Attenuated Total Reflectance) you measure the reflection the light over your sample and a reflective crystal like ZnSe. It depends on the dispositive for measure you use.
The sum of the light transmitted , reflected , and absorbed is equal to the amount of incoming light. Depending on sample that you would like to characterize you can choose a proper FTIR configuration in terms of sample positioning and detectors. In particular, for samples IR opaque (trasmittance = 0) you can use their reflectance to get informations. This means that you can collect specular reflection and then by Kramers-Kronig relations calculate transmittance. Take into account also the existance of integrating sphere, as FTIR accessory, that can collect diffuse component of infrared light to fully characterize the reflection properties of samples.