Cannot decompress content stream of PDF specifications

The content stream of the first page of the PDF Specification contains the word stream, followed by a CR LF, followed by the bytes: 9A FC 2C 16 E1 DB 83 80 92 08 82 AD 7A A1 .... The filter is FlateDecode. The stream object is at position 0x4312.

I can't decode it. I get a few bytes of nonsense, and then it throws an exception. I have tried:

• .NET's DeflateStream
• DotNetZip's DeflateStream
• DotNetZip's Zlib Stream
• My LZW decoder (tested elsewhere)
• My DeflateStream (tested elsewhere)
• Skipping the first 2 bytes, as suggested elsewhere

The PDF opens in Acrobat. Acrobat Preflight shows that the content stream begins with 'BT'.