The number PCR amplicons has an exponential grow with a slow phase at the beginning and a saturation slowdown at the end. The start of the exponential grow depends on the initial DNA concentration, so if you have a lot of initial DNA the exponential phase starts earlier so less cycles are needed to saturate the reaction. If you have few initial DNA you will need more cycles to reach the exponential phase. The final slowdown happens always when the polymerase activity decreases and the dntps and primers are depleted. So less cycles for high initial DNA concentration and more cycles for low initial DNA concentration
It can vary depending on when the program was designed. Older protocols with less efficient polymerase or on older machines often say to use more cycles. Newer equipment and enzymes are more efficient and newer protocols will recommend fewer cycles. It's pretty much a "we did it this way and it works so why change it" mindset.