In a molecular phylogeny of fishes produced using a cytb marker of 704 bp, Sota et al. (2005) (http://www.ncbi.nlm.nih.gov/pubmed/15684588) calibrated an ML clock tree with a node corresponding to the MRCA of two lineages that are assumed to have diverged in allopatry for 3.5 million years.
The 'node height' of this calibrated node is 0.047 (in Fig. 3, illustrating the ML clock tree, 'node height' apparently = number of substitutions per site).
The authors state that the calibration resulted in a "substitution rate of 2.7% per million years". Later on, they state that "3.5 million years corresponds to 9.4% sequence difference, giving a molecular clock of 2.7% per My".
I suppose that: 9.4/3.5 = 2.7 ...
The node height (0.047) should in fact be the branch length, or the number of substitutions separating the MRCA to one of the two sister lineages, divided by the length of the sequence (704), that is, the (average) number of substitutions per site. In this case, the 'divergence' between the two sister sequences should be twice this amount (the number of substitutions per site between the two sequences, along both branches), or 0.094.
By dividing the divergence (0.094 or 9.4/100, or '9.4%') by 3.4 million years, the authors found a 'divergence rate' of 2.7% per million year.
This however is referred to as the "molecular clock", or the "substitution rate".
Indeed, many authors (including me) would in this case use the term 'substitution rate' to indicate the average number of substitutions per site between the MRCA and one of its descendants, that is 0.047/3.5 = 0.0134 per million year, or '1.3% per million year'.
(incidentally, it always puzzled me why this complication of the '%', which should correspond to a 'rate per 100 million years').
When Sota et al. (2005) compare their "fish cytb molecular clock" of 2.7% per million year with the estimates of different studies (Orti et al. 1994; Cantatore et al. 1994), they find a range 0.8-2.8% per million year that is perfectly compatible with both the 'divergence rate' (2.7%) and the 'substitution rate' (1.3%) calculated above ... a misunderstanding of these rates is obviously very easy, since it is entirely possible that these other authors reported 'substitution rates', and not 'divergence rates'.
I'd be happy to share your thoughts about this topic.
Gianluca