Eric,
(Edit: after looking at the equation you linked, it looks like it accounts for the spectral window and the discrepancy is probably just the spur reject setting, leaving the full reply)
A couple things that come to mind regarding your measurement discrepancy. The SA44B has spur reject processing enabled by default. Unless you either disable this in the settings file menu or lower your VBW substantially, this will contribute to a lot of your error. Additionally you will want to account for the window bandwidth. All spectrum analyzers will utilize a spectral window. In our software this is the RBW shape control. By default, we use the flat top window. The spectral windows will also contribute to the calculation. Below I linked another thread which goes into detail regarding this and also you will find a code snippet in one of my replies which has the channel power code calculation from Spike. You can compare your code to it.
The spectral window is also why you are seeing a different sweep size than you expect. When using a spectral window, larger FFT sizes are required to achieve the same RBW, thus the larger sweep size. If you are unfamiliar with these window functions, see the other link below.
If you have follow up questions, please let me know.
https://en.wikipedia.org/wiki/Window_function
Regards,
Andrew