The recommended minimum number of transitions you would need to be confident that you have correctly identified your peptide depends on how much other information you know about the peptide.
Do you know the approximate retention time of the peptide?
Do you have a spectral library and therefore know which transitions are likely to have strong signal?
If you have no other information about your peptide, then I would recommend collecting data for all of the y-ions of the peptide and there is a reasonable chance that Skyline will choose the correct peak, but the probability of getting the correct answer might still be below 50%.
If you have a spectral library then I think we often recommend choosing 5 ions.
You might find it helpful to take a look at the Targeted Method Refinement tutorial.
Transitions that do not appear to be coeluting have interference and should not be used for quantification. If you use Skyline-Daily you can right-click on the transition in the Targets tree and uncheck the "Quantitative" menu item, so that their peak areas will not be included in values such as "Total Area". Non-quantitative transitions still get used in terms of Skyline deciding which candidate peak is the correct one.
(In version 19.1 of Skyline, marking transitions as non-quantitative was a little more complicated and involved using the Document Grid).
If your heavy peptides use Deuterium then it is important that your peak boundaries cover the entirety of both your heavy and light signal. This is because Deuterium causes a retention time difference between heavy and light and failure to include all of either the heavy or light will bias the ratio value that gets calculated.
If you have not used Deuterium then, in theory, the retention times of the heavy and light peptides are almost identical and therefore truncating a peak will not bias the ratio. You should choose peak boundaries so as to minimize interference from other molecules.