The BICEP2 paper reports a tensor/scalar ratio r=0.20+0.07−0.05, but then says:
This is the value taken before corrections. There exist contributions to the B- mode due to changes in the photon polarization of the CMB while it is traveling before reaching the detector. The dust is the interstellar dust that has to be modeled.
Subtracting the various dust models and re-deriving the r constraint still results in high significance of detection. For the model which is perhaps the most likely to be close to reality (DDM2 cross) the maximum likelihood value shifts to r=0.16+0.06−0.05
Which value is proper to compare with Planck's r<0.11?
Assuming that Planck has corrected for the dust before giving its limit, ( a reasonable assumption since the existence of dust is known) it is the second value you have to compare with, which is different only by 1 sigma from the bound given by Planck, i.e. is consistent.
At these errors one does not call measurements definitive. Once many standard deviations separate old from new measurements, old ones can be ignored.