Removing non-negative (non-imaginary) frequencies is impossible within optimization. Indeed, those imaginary frequencies associated with free torsional movements (a common and good example is methyl group) within optimization of reactant(s), product(s), and intermediates as well as unwanted imaginary frequencies within optimization of saddle points must be tried to remove. In this sense, you can employ several techniques depending your approach. For instance, rotation that group to which imaginary frequency is associated and, re-optimization may resolve the problem. On the other hand, inclusion "opt=tight" together with "scf=tight" along optimization and "int=ultrafine" keywords within frequency calculation mostly leads to correct unexpected imaginary frequencies. As you well know, contribution of imaginary frequencies among partition functions is automatically ignored by Gaussian when thermochemical parameters are going to be calculated. In consequence if unexpected (unwanted) imaginary frequencies are not corrected, your thermochemical functions include some serious problem and they are not reliable.
While I have not the Truhlar’s article you mentioned within your question, I guess it should be a manually calculation. Presumably, entropy is not extracted from the Gaussian output and, instead, it is computed manually in such a manner that contribution of low frequencies is neglected/improved/corrected.
While I hope my comments are found helpful, the more reliable comments of other researchers who are more expert than me will surely help you appropriately.
It's true that you cannot avoid this frequencies in the optimization. There are two approximations to correct this little frequencies (one done by Truhlar, and other by Grimme). The Truhlar approximation is based on setting all the frequencies lower than a cut-off value to this cut-off. The Grimme approximation is based on using the rigid rotor approximation for the lower frequencies and the armonic one to the rest (using a dumping function to avoid technical problems).
I developed a free code to do these corrections automatically (with Prof. Robert Paton in University of Oxford). In addition you can change the temperature, concentration and other features in the thermochemistry calculation. Here is the website:
https://zenodo.org/record/60811#.WNKLXUfau90
You can cite it using the associated DOI in the webpage.
just a follow-up on your answer. I used your script and it works very well!
I have just a question: I would like to evaluate the impact of eliminating rotational and translational contribution on entropy from the final Gibbs free energy. Is your script affecting only the vibrational contribution to entropy, isn't it?
The program actually computes every single term of the entropy and the entalphy. The question is that if you are applying quasiharmonic corrections, only the vibrational and the zpe change respect the Gaussian output. If you want to modify or exclude the rotational or translational entropy, you can modify the program easily (I don't have the program now, because I'm on the phone), commenting the terms of entropy calculation or modifying the last sum of terms. If you have problems, let me now and I would do that.