They are more stable than the C18:x, C20:x, C22:x, highly unsaturated FAMES, so your losses are probably due to either volatility or injector discrimination.
You can check for discrimination by using one of the standard AOCS FAME oil test mixes, or prepare you own. If you are getting higher unsaturated and long loner oils, I'd want to ensure that the other lab didn't have analytical problems by using a reference oil or FAME mix, as it's more usual that the unsaturates are lost due to column activity etc.
EPO should be easy, I did 100s of samples in the 1990s, and I derivatised using the quick procedure of Bannon ( it's an ambient temperature methanolic KOH method which takes about 6 minutes a sample, and you can do about six at a time in small 4ml screw cap vials - I think we used C19:0 as the Int std ). Bannon later reported errors can be introduced by the method, but I found that really good mixing ( vortex mixer ) gave quick and accurate results for most oils, including several marine oils.
These days, I suspect there are standard methods for EPO, especially as the oil is sold based on Gamma Linolenic Acid content. I'd look for one of those, or use a second method to check your derivatisation.
The AOCS marine oil method should work for most marine oils, remembering the exclusions listed in the scope of the method.
Please keep having fun,
Bruce Hamilton
