For any instrument that quantifies via integration of a peak (GC, NMR, etc.), sample concentration (peak height) is important especially with low signal/noise ratios. Obviously if the peak isn't in the sweet spot - so to speak - (where its area is significantly above the baseline, but not too big that it saturates the detector) the quantification may be unreliable. Also peak shape is a consideration in integration. If the peak is too broad, much of the area under it can be obscured by the baseline. On GC/MS or GC/FID in particular I used to have problems with peaks 'tailing', that is a broad lingering asymmetry after the peak max which would often overlap with species with greater retention times. No es bueno for quantification. Baking out the column, finding a better heating profile/method usually did the trick.
Optimal methods are going to vary widely based the instrument you use, the solvent, what components you have in your mixture, there relative amounts and concentration etc.. A lot of trial and error. As a general rule though typically when I had 2 components that were close in rt, but wouldn't resolve, I'd move to a method with a much slower ramp for the column temp, upping the carrier gas if possible on your equipment might help, too.