Computation of vibrational frequencies

Hello everybody,
I would like to have some hints to perform accurately the computation of the IR spectrum of coumarin-3-carboxylic acid (CAS: 531-81-7).

I used the atomic coordinates determined from X-ray diffraction data to draw the molecular model, but as I computed the IR spectrum I obtained peak values very far from experimental data derived from the literature. I also tried to use two neighbouring molecules forming the dimer, but to no avail.

I tried to optimize the geometry of the single molecule and of the dimer by Forcite and DMol3, to no avail.

How can I improve IR spectrum computations?
Thank you very much in advance

LAV