Since the software maker always do benchmarking before the release I believe that any software is pretty much correct, the one that makes the different is the method (the theory level) and the basis set that you're using.
I am a vibrational spectroscopist, as for as my experience is concerned, I am working with molvib for about 6 years and it will give you accurate results. But the thing is you have learn the algorithm to work with molvib and it may take months. Instead you can use veda and gar2ped they are simple and as simplicity enters accuracy decreases.
In case if you have any queries let me know and am ready to help you in any regard.