Thanks for the advice. I guess I should have checked my computations a billion plus one times <g>. I found my error, a simple sign flip. When corrected, the prediction agrees very well with the observed. It should, not only because all the theory is correct, but the actual values of the rest mass of the electron, proton, etc., used those exact observed values for their update. This is the way NIST does things, which is about the only way to do it. But then it really doesn't make good "scientific method" sense to predict the value when it was used in the update. That had better work, and it does quite well. The best thing to do is get a completely independently derived observed data set, and see how the theory performs. I have one, and I'll try to do this independent comparison.
The usual basic Dirac equation, without QED corrections is given as
[math]{E_n} = \mu {c^2}{\left[ {1 + \frac{{{\alpha ^2}}}{{{{\left[ {n - j - 1/2 + \sqrt {{{(j + 1/2)}^2} - {\alpha ^2}} } \right]}^2}}}} \right]^{ - 1/2}} - \mu {c^2}[/math]
This is given on p. 607 of Merzbacher's Quantum Mechanics, 2nd. ed. The positive first term is what Merzbacher gives, but if you follow his derivation, the total orbital energy (negative valued for bound orbitals) needs the [math] - \mu {c^2}[/math] term. For spherical orbits, j = 1/2, so that
[math]{E_n} = \mu {c^2}{\left[ {1 + \frac{{{\alpha ^2}}}{{{{\left[ {n - 1 + \sqrt {1 - {\alpha ^2}} } \right]}^2}}}} \right]^{ - 1/2}} - \mu {c^2}[/math]
This is the equation I used to do the predictions. The simple error I made was to the subtract a QED correction when it was suppose to be added, and visa versa. You can see the kind of "mislogic" in comparing this prediction to an observed "input" data point (the 1S1/2 - 2S1/2 value, for example). This data point (and many others) were used to "update" the values of the electron and proton masses, etc. Those values therefore "parameterize" the Dirac theory to conceptually "work perfectly." That's ok, this is really the best way to do it, assuming the theory is correct.
But is the theory absolutely correct? To question Dirac theory is blasphemous, but there are some aspects of it that confuse me. Consider hydrogen's spherical orbitals. Without special relativity, the Bohr theory results in simple quantized orbital radii and velocities. Of course, the rest mass of the electron is in these equations. When special relativity is introduced, the electron mass "on or in" the orbital time dilates and is bigger. The relativistic increase in the electron's mass increases the total orbital energy, resulting in the above equations. Do the orbital radii decrease and the velocities increase? The probability theory would suggest they do, because the ground state radial pdf, for example, has a mean and mode closer to the proton after relativistic effects are introduced.