Well let me answer this question with respect to the negative imaginary frequencies in calculating the IR spectra in Gaussian. There are a couple of factors you need to consider. If you get too many negative frequencies for a small size molecule then it implies you have not reached the global minimum. Probably you have reached some local minimum for the optimization procedure. Feed the values in Chemcraft and plot the optimization curve, you'll get an idea of the stability of the procedure. Ofcourse you have to further optimize the structure. Another case is the optimization of a big cluster. In this case if you get a single negative frequency with value close to zero, it means it is simply a transition state. You may choose to optimize further to reach the global minimum.
In that case, should we take that minimum energy conformer (which has a negative value) and optimize it using DFT, 6-311++G(d,p) and then take the PES scan again?