Firstly, the 4% perfect use failure rate is merely someone's guess, particularly James Trussell as author of "Contraceptive failure in the United States", Contraception vol. 70, pages 89-96:
Our estimate of the proportion becoming pregnant during a year of perfect use of withdrawal is an educated guess based on the reasoning that pregnancy resulting from pre-ejaculatory fluid is unlikely [citations to two 1992 articles]
On the surface, the two studies cited in the question sound contradictory, but it is important to consider the details, especially the experimental methods. Because the 2011 study cites the early study and gives reasoning for the differing results, it should be considered to supersede the earlier study.
The reason why our study was able to demonstrate motile sperm in pre-ejaculatory fluid whereas other studies have failed to do so might lie in the promptness with which we examined the samples. In our IVF unit, the room where men are able to produce their samples is immediately adjacent to the laboratory. We briefed our volunteers appropriately and arranged for an embryologist to be positioned at the microscope awaiting each sample, and we are confident that samples were examined within 2 min of production. After this time, low volume samples can dry out and microscopic examination becomes extremely difficult.
and
the actual number of sperm in the pre-ejaculates was very low.
Furthermore:
It would appear from our study that some men repeatedly leak sperm in their pre-ejaculatory fluid while others do not.
Also, the earlier study asked men to abstain from ejaculation for 3 days. Obviously immediately after a man ejaculates there is still sperm in the urethra. Since most men are ready and able again in much less than 3 days, maybe even an hour or less, there is risk of sperm being present from a previous ejaculation. According to Longevity of spermatozoa in the post-ejaculatory urine of fertile men, motile sperm remains detectable in urine for several hours after ejaculation.
In conclusion, the state of knowledge is merely that there is some non-zero probability of pregnancy for reasons other than ejaculation, plus there is difficulty for a scientist to distinguish between perfect and not perfect use for withdrawal, particularly:
- There can be sperm still in the urethra from a previous ejaculation.
- Some men leak sperm into pre-ejaculatory fluid ("We are unable to say how this finding might translate into the chances of pregnancy if these samples of pre-ejaculate were deposited in the vagina except that the chances would not be zero" according to the 2011 article)
- "Men may be unable to predict the moment of their ejaculation and subsequently fail to admit to this" (2011 article)