When you draw a calibration curve is always the slope is higher than the original point at its intersection with the y-axis, with the knowledge that the solution used is blank solution.
and when to get on the value of intercept the least possible?
please check and verify my previous speakers' hints and advices. Simply measure a blank sample (same solvent etc, but no analyte). You will most likely see a response leading to an offset in your linear regression model although no analyte is present. This is due to the chosen analytical method.
I agree with your statements and I will include another important point: The charging process of the electric double layer is very important. Why we have not exactly zero of current even in an ultrapure electrolyte? This happens because the charging/discharging process of the electrode is happening and this must be taken in account.
In addition, this is the reason why CV is not the best technique for quantification of analytes at low concentrations. I loved the commentary of Prof. Markus about the importance of take in account the "blank" to determine LOD and LOQ. The majority of the papers make calculations without consider the existence of noises.