Standard Plate Count

We are conducting a trial of a process on liquid full cream milk. Control and trial samples are held for analysis on a two day continuous cycle with an unopened sample used for each test. One of the tests is Standard Plate Count. On review of the results we do not understand why there are such wide variances in results between samples tested For example;

Result one in test/date order (cfu's)
Each successive result is milk two days older than the previous test. The testing facility is an accredited milk analysis laboratory.

7,000
600,000
3,000
2,200,000
160,000
41,000
15,000,000


What explanations could be applied to these results?