the lowest copy number sample points impacted faster and sronger, then you just need to stabilize by adding a neutral nucleic acid background in your standard curve. I usually use water containig 10ng/microl yeast tRNA to perform serial dilutions. This will first create a reaction background similar to your RTQPCR reaction, but also stabilize your DNA copies; I can freeze and thaw (min 20C) more than 50 times the same standard curve sample without any loss in Cts, from 10E6 to 10E2 copies. When I tested the same standard curve but diluted in water only, then the 10E2 started to be slightly affected after one freeze and thaw and then crashed further; then higher copy numbers samples were also affected after 2 to 3 freeze and thaw.