As William mentioned, LOQ or LOD are determined with a different approach (e.g. s/n ratio) compared to linearity which is a relation between signal and concentration. LOQ/LOD gives you an idea of the detection sensitivity. In some cases, LOQ falls inside the calibration curve, but in other cases, LOQ can be lower than the lower limit of the calibration curve.
@ William @ Marcos: Thanks for the reply. For LOD determination using 3.3SD/Slope formula, the SD and Slope is the standard deviation of the area under linearity curve and slope of the linearity calibration curve... Is it correct?
LoD can definitely fall outside the linearity. According to FDA guidelines, LoQ (they use the term lower limit of quantification, LLOQ, instead) must fall within the linearity. (linearity is the range between LLOQ and ULOQ).