There are a few reasons of misindexing which need to be distinguished. One kind is caused by the software (too strong or weak restricts from the indexing (band number, reflector number, Hough resolution etc.)), one is the result of the sample (e.g. really existing (nano)twins, bad patterns (fake bands) ), one is related to the phase (the selected reflectors are not that dominant as the software believes (mainly caused by a pseudosymmetric (hexagonal) layers parallel to {111} with a symmetry of 6mm), and sometimes simply the operator is not experienced enough. For simple phases like bcc and fcc all systems are quite stable during indexing and such problems do not occur anymore... The major problem is the operator who believed the advertisement of EBSD vendors about the reachable speed. In case of occurring misindexings I would recommend to improve the pattern quality. If they are OK, you should think about the projection center. A post-processed "symmetrization" of data applying "fake" symmetry elements might change also really existing twins or other orientation relationships. I realized this during presentations where these tools have been presented from the manufacturers. I call this data manipulation. If you measure EDS you also cannot (or should not) remove disturbing peaks which are the result of sample topography or other artifacts. Then you explain them or you simply measure the same sample again taking care about this problem. Usually it is the insufficient pattern quality.

Regarding your question: I only know a paper from Luke Brewer and Joseph Michael who discussed the risk of data cleaning in general.

Gert, thank you very much for the detailed explanation.

Now my question about a publication seems to be not actual any more.