The pattern labeled "8-bit" in the page from fts_17_1704.pdf looks OK except
for the fill and stroke labeled '8-bit'. Both objects 25 and 50, which appear
to be the image source for the pattern are JPXDecode filtered and have
/SMaskInData 1. For example, the -dPDFDEBUG output:

%Resolving: [27 0]
<<
/BBox [
0 0 72 96 ]
/Filter /FlateDecode /Length 32 /Matrix [
1 0 0 1 135 535 ]
/PaintType 1 /PatternType 1 /Resources 24 0 R
/TilingType 1 /Type /Pattern /XStep 72 /YStep 96 >>
stream
%FilePosition: 9761769
endobj
%Resolving: [24 0]
<<
/XObject 26 0 R
>>
endobj
%Resolving: [26 0]
<<
/Im0 25 0 R
>>
endobj

%Resolving: [25 0]
<<
/Filter /JPXDecode /Height 1600 /Length 3076639 /Name /X /SMaskInData 1 /Subtype /Image /Type /XObject /Width 1200 >>
stream

--- then later in the processing ---
%Resolving: [25 0]
%Pattern: << /Type /Pattern /BBox [0 0 72 96] /TilingType 1 /PaintType 1 /.pattern_uses_transparency false /PaintProc {<< /XObject {26 0 resolveR} >> .pdfpaintproc} /XStep 72 /Resources {24 0 resolveR} /YStep 96 /PatternType 1 /Matrix [1 0 0 1 135 535] /File (x\234+\34457R0\000BK30\225\234\313\253\357\231k\240\340\222\317\033\310\013\000Wd\006\007) /Filter /FlateDecode /Length 32 >>
120 150 m
160 270 200 30 240 150 c
200 270 160 30 120 150 c
f*
%Begin PaintProc<< /XObject {26 0 resolveR} >>
q
72 0 0 96 0 0 cm
/Im0 Do
%Resolving: [26 0]
%Resolving: [25 0]
[(jP  ) 4]
[(ftyp) 20]
[(rreq) 173]
[(jp2h) 557288]
[(ihdr) 14]
[(colr) 7]
[(colr) 557171]
[(cdef) 38]
[(res ) 18]

I assume that the arrays printed after %Resolving [25 0] are from the JPX
decode, but don't know.

Since this looks like an area that Michael is working on, assigning to him.
(sorry, mvrhel)