This is what I found on a poster from a former colleague; it is a nearfield IR spectrum (scattering-type SNOM) corresponding approximately to the reflection, comparing anatase with as-deposited.
In FTIR reflectance mode, as well as in KBr tablets (see N.Chukanov, Infrared spectra of mineral species), it is impossible to distinguish anatase and rutile. It is possible in transmittance via thin (a few microns) plates. But it is much easier to use Raman