1. There are two ways I'd approach this.
First, research which molecular mass is for which version of the substance. Sometimes it is the difference between one type of Vitamin A vs. another.
If they are 3 measurements for the same exact substance, you want to then go with the one that is in Monoisotopic Molecular Weight (g/mol) in case another is using a different scale.
Furthermore, you may wish to go with the one from a site that is more trusted over another if you got 3 values from 3 different sites.
However, in the end, you could make this even simpler.
Second, add all three into a frequency program for said substance and use all three.
Even at 3 minutes each, you only have 9 minutes there. Also, you never really want to have just one frequency loaded which would repeat as a static frequency. So this will give you a bit of variation if you are going to run just this one substance.
2. Also, those weights are so close to each other, the difference in calculated frequency will be very minimal. May not even be significant enough due to resonation margins. May only be significant enough when reducing the frequency to a very very low sub harmonic.
For more details, please check the link: