@pnelson should look at this example closely. It shows convergence at a functional level, convergence at a gene level (vis vi: Winston Ewert: The Dependency Graph of Life), but divergence at a sequence level. Is that right @davecarlson?
FIgure 3 appears to be the key data:
A) Bayesian inference of coding sequences of the Petunia AN2 gene. H1 and H11 encode functional proteins, whereas H2–H10 are pseudogenes.
(B) Graphical display of the coding region of AN2 ordered into the eleven haplotypes (H) classified according to the nature and position of the inactivating mutations. ATG, start codon: nt380: nucleotide 380; inverted triangle, indel, with number above indicating number of bases deleted or inserted; red star, stop codon.
(C) Partial nucleotide and translated protein sequences of the short-tube species P. inflata (H1), P. axillaris / P. exserta (H10), and P. secreta (H11). The 1-bp deletion in P. axillaris / P. exserta translates into a truncated protein. A compensatory 2-bp deletion leads to the functional AN2 of P. secreta .
(D) Introduction and silencing of the P. secreta AN2 gene modify floral color (from left to right): P. axillaris; transgenic P. axillaris :: AN2 P. secreta; P. secreta; P. secreta AN2-VIGS; note the white spots on the purple flower. Starred accessions are based on the coding sequence, see Table S1.