non-linearity with nested standards

[mav]

i am running hplc analysis with the samples bracketed by a set of 4 standards. a recurring issue has been that the curve before the samples looks great on its own, and the curve after the samples also looks great on its own. however, standards on the higher end are simply not reproducible, which of course throws my whole data set off and the linearity decreases dramatically when using all points. i have also seen a similar problem with some gc analyses. i would not assume that there are any problems with the dilutions of the standards since we are doing everything according to GLP (class A pipets, being very careful, etc.). any suggestions on how to improve this? i assume it must be an instrument issue, but don't know where to start.

[Kwet]

Could you give more details?
What is your detector?
Could you give concentration levels? Which one failed?
Does the bad results comes from the sample table position or concentration level?
Have you considered the higher level is upper the linearity range of the detector?

[mav]

Here are the details. I am running Carbamate analysis, so I am using a gradient method (water, MeOH, and ACN are the mobile phases) and post-column derivatization and fluorescence detection. The post-column reaction is at 80 degrees F and the post-column solvents are 0.05N NaOH and OPA. Column is thermostated at 30 degrees and the flow rate through the column is 1.2 ml/min. My standards are 5ppb, 50ppb, 200ppb, 350ppb, and 500ppb. Typically, the 350ppb and 500ppb are not reproducible before and after the samples. (Could matrix effects play a role?) Because the before and after curves have great linearity on their own, I assume that the linearity range of the detector is not a factor, but perhaps some other instrumentation issue is the culprit.