NPN needs deep-nwell process usually is expensive. so from cost point of view it is not used very often.

You have to use, what is available on the technology you use. CMOS technologies rarely have optimized diodes, or BJTs. These devices are formed by process steps required for CMOS transistor formation, and they are optimized for MOS transistors. One can say that the BJTs in a CMOS process are often parasitic devices. As /aznsj said, PNPs are usually better suited in a general CMOS process.