I suggest diluting the samples with deionized water. I had a similar problem testing for total cyanides many years ago using a colorimetric test and noticed that samples from a particular industry were reading less than the blank. I figured there was some type of interference and so diluted those samples 1 to 10, 1 to 50, 1 to 100 and 1 to 200 with deionized water and ran them again. Low and behold I calculated that those samples actually contained around 70ppm of total cyanides taking into account the dilution factor. This was later confirmed by the local munucipal water treatement facility and the industry responsible had to change their process. This will of course raise your limits of detection. Alternatively you may want to look into ion chromatography.

Thanks for your response, Thomas.

Unfortunately I am already at the limits of dilution and other detection methods are outside my lab's capabilities. However, I think I have stumbled across a possible solution. Adding manganous ions (Mn2+) results in formation of a brown precipitate which is presumable a sulphide. When this is filtered out, the clear solution responds to the ferric ion addition with normal color development. There does appear to be a minor interference in the reading, but I think I can compensate for this with an appropriate blank solution and also by optimising the dose of managanous ion I put in. I'll be sending some check samples off to a commercial lab to make sure it is working.

That sounds plausable. My suggestion would be to spike some aliquots of a sample with ferric ion before and some after the addition of the Mn2+ for comparison of recoveries. Best of luck. If it works that could be something that is potentially publishable.

You may use a mild oxidant to get rid of your reducing agent. Iodine may work. Do not add a great excess.