I became intrigued in this paper because it was pointed out to me that Phytophthora sojae is listed in Figure 1C and this is a genome we have been interested in. A quick blastn search of the JGI P. sojae portal identified some problem sequences. For example, there is a problem with the last ~1.3 kb of scaffold_632. This part of scaffold_632 matches 100% a human sequence. However, this sequence is not present in the GenBank accession of this genome (Phytophthora sojae strain P6497, whole genome shotgun sequence, AAQY00000000) and was presumably filtered out.

Conclusion: As indicated in the other comments, the problem has been overstated. In the case of the P. sojae genome, it seems that NCBI contamination screening pipeline has worked.