Dear Eric,

This paper titled "Guidelines for DNA taxonomy, with a focus on the meiofauna" by 

D Fontaneto, JF Flot, CQ Tang - Marine Biodiversity, 2014 - Springer may help you out!

Good luck!

This is the tool you want to try and this does exactly you want: https://github.com/TGAC/kontaminant

idea behind this tool is to target the problems like mixing of samples or symbionts. Kontaminant clusters the reads based on the reference species. Please let me know if you have difficulties in downloading the tool.

Best wishes

Shab

deconseq would be a good tool for your task if you just wanna remove a certain  organism

If you are using RAD-seq I would suggest looking at Stacks. Stacks can demultiplex and cluster your RAD-tags and creates output files for alignment to a genome and gives suggests on how to go about do this. It is pretty easy to use and does a good job with clustering reads and useful output files.

http://catchenlab.life.illinois.edu/stacks/manual/

Hope this helps

BBMap will allow very low alignment thresholds which can be useful. The tool suite also has a set of kmer comparisons that allow one to tag reads that share kmers with a reference.

But, I think an easier approach is to not filter the raw reads but to construct the set of reference loci using Stacks or pyRAD and then more aggressively search that much smaller set of sequences for contamination using Blast to a non-redundant database. 

In the end though, unless you have a good positive selection using a reference genome of what you want you will be left with many loci of indeterminate origin. It is crucial to have one "clean" sample of the species you want to genotype and use those reads to positively ID the loci in your reference set.