The basic idea behind a practical approach to a computational algorithm for the properties of the upper atmosphere is that we first compute the solution to the complex equations all the way from 86 km to 1000 km at fairly tight intervals. This requires a considerable amount of code and computation, and you may examine this procedure in the program ussa1976.dpr.
To make a practical computing scheme, one selects a subset of the full output of ussa1976 that covers the range of 86 to 1000 km and finds intermediate values by interpolation. I spent a good bit of time experimenting with the number and choice of data points. I also concluded that one should use the logarithm of density and the logarithm of density vs. altitude as these quantities are almost linear with altitude. The temperatures are computed from equations taken from the reference publications You may examine Pietrobon's original program uasa2000.dpr (in Pascal) amd my subsequent revision atmos76.f90 (in Fortran).
I am open to the idea of revising the set of points used for the interpolation of pressure and density.