More about Progesterone Testing and the Value of Pooled Progesterone Testing for Luteal Phase Defect

"Single or multiple pooled serum progesterone values have been used between 21-24 days into the menstrual cycle. A progesterone level greater than 10 ng/mL may represent adequate progesterone support from the corpus luteum. Most authors now agree that a single mid luteal phase progesterone level is not sufficient to evaluate the adequacy of the luteal phase."
From: Lutenizing Hormone Deficiency

"Among women who present with a history of infertility, approximately 3-5% of those with infertility have a true luteal phase defect, with repeated cycles lacking adequate luteal support. The incidence of luteinizing hormone deficiency is higher (10-15%) in subsets of patients who are infertile and have had [repeated miscarriages]."
From: Lutenizing Hormone Deficiency

"As with many hormones, progesterone levels fluctuate hour to hour and day to day, so that a single level is not always an accurate reflection of ovarian function. Also, it is important that progesterone levels are sufficiently elevated for an adequate number of days after ovulation. For these reasons, it is ideal to check more than one blood progesterone level during the time in your cycle when progesterone levels are the highest, approximately 5 to 9 days after the LH surge. In the pooled progesterone test, blood is drawn on three days between the 5th and 9th days of the luteal phase. Blood samples will be "pooled" together, and a single progesterone level will be determined, representing an average of your progesterone levels."
From Pooled Progesterone Test (PDF)

"The results suggest that the most sensitive and specific prediction of low integrated progesterone (and therefore of LPD) was a value of <30 ng/mL for the sum of three (midluteal) serum progesterone measurements (100% sensitivity, 80% specificity).

Even a single (midluteal) progesterone result of <10 ng/mL was highly predictive of LPD (84% sensitivity, 82% specificity). However, a diagnosis based on a single progesterone determination continues to be controversial. Progesterone is secreted in a pulsatile manner, and serial samples obtained over a 24-hour period can range from 2.3 to 40 ng/mL.3 Such fluctuations raise questions about the reliability of a single progesterone determination for the differential diagnosis of LPD. The pulsatile secretion of progesterone argues for either assaying multiple specimens separately or pooling multiple specimens for processing in a single assay."
From: Progesterone: Physiology and Clinical Utility (PDF)

With a pooled progesterone test you should do it fasting, on any three days during your mid-luteal phase -- 5-9 dpo.

I've also read that luteal phase dysfunction is best tested in tandem with an endometrial biopsy and basal charting.