if you looking for trace analysis..you may preferably prepare in low range concentration. however some compound may less sensitive to your instruments. what more important is your linearity given good r2 > 0.990
It depends on your detection method and instrument sensitivity. You should try some of your standards and find the concentration range where the instrument response is linear (r2>.95) and then work in that range.
For selecting the calibration range, you should know the target concentration of the compound, I think the best calibration range is ca +/- 20 % the the target concentrations. If you do not know the target concentrations you can try to develop wider calibration range than you can estimate the concentration of the target compound; after this than you can develop calibration range +/- 20 % estimated concentration to determine the exact concentration. To express the linearity of the calibration curve, you cannot use "r" as sole parameter, other parameter such as Vxo, Mandel Test, F ratio etc must be determined. Please read e.g. P. Araujo, J. Chromatogr. B. 877, 2224-2234, or our chapters regarding Validation in: HG Brittain (Ed.) Profiles of Drugs Substances, Excipients and Related Methodology, Vol. 32 (Validation method of analysis by using Chromatography) and 37 (Validation method of Analysis-Updated 2011), Elsevier Academic Press, San Diego, New York