cummeRbund diffData() or fpkm() values

[jeales]

I'm getting different FPKM values from diffData() and fpkm() on a couple of different isoform IDs

> diffData(getFeatures(cuff,"TCONS_00246566"))
      isoform_id   sample_1 sample_2  status value_1 value_2 log2_fold_change test_stat p_value q_value significant
1 TCONS_00246566 macrophage monocyte LOWDATA       0       0    -1.79769e+308         0       1       1          no

diffData() reports an fpkm of 0 for sample_2 (reported as "value_2")

> fpkm(getFeatures(cuff,"TCONS_00246566"))
      isoform_id sample_name      fpkm conf_hi conf_lo quant_status    stdev
1 TCONS_00246566  macrophage 0.0000000  0.0000       0      LOWDATA  0.00000
2 TCONS_00246566    monocyte 0.0972739 49.0851       0      LOWDATA 24.49391

but fpkm() reports an fpkm of ~0.09 for sample_2

I know that csHeatmap() uses fpkm() to extract values for plotting, but which is the most accurate value?

I thought cuffdiff values (available from diffData()) were supposed to be the most accurate?