Hello, several reasons may cause this problem.

1. Contamination may be your sodium chloride contain potassium, water, HCl etc.

2. Problems with calibration curve

If you have curve like in picture or similar something measurement concentration of point 2 you will receive difference from real concentration equal too absorption d1 in picture if point 3 difference d2.

It means there is no reprocibility in your measurements when you make replicate measurements. The problem must be with sample preparation method and in particular dilution factor. There may be a problproblem with detector. It is to be conditioned to be responsive reproducible.

There could be several problems.

1. Your solution may contain particles which the AA detector will see, just like a spec. Thus, filter your water before use.

2. Be sure you acidify your sample and standard (1 drop per liter with Nitric Acid will do) to ensure all the Potassium atoms are in the same valence state (no metal).

3. Potassium can be determined both by absorption and emission at 766 nm (like any Alkali earth element, Na, Li, Cs...).

Have you tried the emission mode to measure K in the F AAS equipment? It works better for K than the absorption mode. Also, I would prepare at least 5 working standard solutions for the calibration curve. Finally, you may need an ionizer suppressor when measuring K, such as La or Cs (for blanks, standards and samples). You may want to have a look at Welz's book on AAS for more info on that. I also have a paper on K determination by flame atomic emission spectrometry, but it is in Portuguese: http://quimicanova.sbq.org.br/detalhe_artigo.asp?id=6844

Good luck!