combine_limit example showcases bad practice?

[andresti]

In the combine_limit example both 1-sigma and 2-sigma uncertainties are added to the expected value, but this seems to go against the CMS Higgs HepData recommendation.
From https://twiki.cern.ch/twiki/bin/viewauth/CMS/HiggsHepData:

For expected limits ("Brazil plots"), please note that it is not recommended to give the 1σ and 2σ bands as independent errors, since the "sum errors" button will sum them in quadrature and give meaningless results. It is better to give the expected limit twice (as two columns/graphs), one with the 1-sigma and the other with the 2-sigma uncertainty.

[AndreasAlbert]

Hi @andresti,
I am not sure I would call this 'bad practice', since the error summing is just a result of the hepdata website visual interface, which is not set in stone and should not be taken for granted. For someone who wants to re-use your data, having 1 and 2 sigma uncertainties on one variable poses absolutely no technical issues. Therefore, I would argue that this is more a question of personal taste at this point.
That being said, I think we (as a community) should come up with actual recommendations on how to handle this and similar "style issues" so that everyone uses compatible conventions.