There is no generally accepted substitution rate. You can calibrate a sutitabele substitution rate for your target marker using geological events or fossil records in BEAST software. Hope this helps.
The best substitution rate that you can use for your study model is a specific one obtained using phylogenetically closely related dated fossils or biogeographic event, even though many have been using a "standard" rate of 2% per million years. It also depends on the species you are considering for the divergence dating. If you want to estimate the divergence of two species phylogenetically distant, you will probably not use the same rates as you would for two cryptic species. This is because the rates that you need to use should account for the apparent "time-dependency" of the rate estimates (see Ho et al 2011: DOI: 10.1111/j.1365-294X.2011.05178.x): the rates should be higher for very close taxa (two populations) and very low for distant taxa (two families). There are much debate around the topic, and one should just decide for themselves on the best strategy of applying the rate and then cross fingers during the review process. Unfortunately, there is no simple answer to your question.