First a disclaimer-- I'm a fruit fly and mouse geneticist so I know about fly eyecolor and mouse fur color...
That being said I have to agree with Philip that it's mostly a crapshoot.
The blue color acts as a recessive with the magenta as a dominant but unfortunately the flower color is controlled by several genes, environmental cues including the pH of the flower (like hydrangea blues and pinks) and a number of genetic elements that jump around unpredictably each type the plant reproduces sexually; all of these affect the expression of the various flower pigment genes.
As far as I understand things a recessive gene
(flavonoid 3′,5′-hydroxylase gene) is needed to modify the purple anthocyanin pigments into blue. Even when a plant has the appropriate recessive genes the degree to which pigments will shift toward blue depends on other genetic elements called promoters that will either enhance (upregulate) or depress (downregulate) expression of the gene for turning purple to blue. Additionally there are co-pigment genes that either enhance the blue or not. Finally, there are these jumping transposable elements that apparently mess with the expression of genes they jump into.
Thus we're back to Philip's answer that its basically a crap shoot.
I would definitely agree with Magda that the selfing would probably be the best bet since obviously the blue making genes and the correct promoters are in the single plants genetic background. the problem is that everytime you do a cross you are setting free a vast set of opportunities for genetic recombination and variation to occur. Variety is not only the spice of life but it is the necessary material for evolution and thus survival.
So the only sure thing is to buy the plant in bloom or buy a division or mericlone of a known blue. If you have OrchidWiz (I don't) it might be helpful to research all possible plants in the cross parentage/history. The more blue the better.
My apologies if this contribution is more confusing than helpful!