If you know what compounds you are hoping to identify you should use those compounds. If you use a calibration curve derived from a different compound to the one you are quantitating, you may have problems.
If you are planning on using internal standards, isotopic analogues are ideal as they have the same chemical properties as your target compound but different spectra. This means that if you use the ratio of the target response to the internal standard response as the y-axis of your calibration curve (as opposed to the absolute response of your target), drifts in response will be compensated for. E.g. if there is a drop in response, the I.S. and target response will drop but the ratio should stay the same.
Did you mean what ions should you use?
you should select the most abundant ion in the spectrum, but if there is a co-eluting compound with the same ion you will need to choose a different ion.