1. Which structure factor amplitude is used in refinement when the labels in the .def file are the following?

labels = "F-obs(+),SIGF-obs(+),F-obs(-),SIGF-obs(-)"

If input file actually has these labels (it should!) then this is what was used in refinement. Note, MTZ file out of any phenix.refine run contains four blocks of information: 1) Copy of original data (Fobs or Iobs, flags, etc); 2) Data actually used in refinement (F-obs-filtered). This may be slightly different from input data by a handful of outliers not used in refinement or if input was Iobs (phenix.refine always uses F, so it will convert Iobs to Fobs using F&W method). 3) Total model structure factors, Fmodel, that includes all scales: Fmodel = k_total * (Fcalc + Fbulk) 4) Various Fourier map coefficients, such as 2mFo-DFc, mFo-DFc, anomalous difference map coefficients. So you can always take MTZ file generated by phenix.refine and know exactly what was used.

2. Is anomalous data used in the refinement when the labels are as follows?

labels = "F,SIGF,DANO,SIGDANO,ISYM"

Yes.

I ask these questions because I am wondering which structure factor amplitudes to deposit,

You need to deposit the original ones, that is from "1)" above. Ideally, "2)" should also go there along with "1)" (but not instead!) because this is what was actually used to obtain reported statistics, such as R-factors.

and I am trying to rationalize differences between R-factors from refinement and Generate Table 1.

Depending which kind of differences you mean (magnitude). Ideally, R-factors should be identical. If it's like 0.1856 vs 0.1858, then it's fine, if larger then some of us here need to investigate what's going on. Pavel