Interestingly Harrison's old code and his book included an interesting coding example for co-ordinate transformations. In the book he published the 3 by 3 matrices for his sample problem. I had checked several of the matrix samples, but not the last one , turns out the code produces a slight variation on the matrix, the published answers agree with mine, but is that because the code is wrong or the book is wrong.
In moving from a 4 by 4 degree of freedom method to a 6 by 6 so I can match the shell elements to the beam elements I took Harrison's code, and then started looking at the final numbers -- I noticed the third beam has some numbers negative when the book says positive.
Turns out that in the book Harrison draws the beam going down the negative z axis - gamma = -90 , but the sample has gamma, but the coded example has gamma = -90