An interesting and meritorious analysis. However, it's impossible to say if codes 13-15 are true codes or just cases of erroneous reporting or factory errors. Certainly, it is a possibility that I wouldn't rule out.

In the case of Miki, I have seen codes beyond those mentioned but again, there is not the distribution you would expect.

Another possibility which I have often considered, is the probability of a 13 letter system. This divides a year into equal segments of 4 weeks and is known to be used by some manufacturers.

The fortnight system is not unknown to Japanese manufacturers. Miyata immediately comes to mind, having introduced what appears to be a fortnight system at the beginning of the 1985 model year. A number of different of different manufacturers have changed formats, which often makes analysis difficult.

I have always stated that my serial number analyses are dynamic in nature and based on current understanding of the data. New data may change views. The more data that is collected, the higher the confidence level. However, they are only educated guesses based on data analysis. The only way to know for sure if an analysis is correct, is via extant manufacturer literature. Unfortunately, there are very few cases of this.