I think おかけで, かつと, and そしで are most plausibly explained as misprints. They don't generally occur, and this novel is too recent (written in 1987) to make comparisons to historical kana orthography.
Although it's possible that バック too is a misprint, another explanation seems plausible to me in this case. In Japanese, a phenomenon called rendaku or sequential voicing occurs primarily in compounds such as ひとびと, in which ひと is joined with another ひと, but the second /h/ is voiced to /b/. This is limited (but not predicted) by Lyman's Law, a set of rules that explain when sequential voicing should not happen.
Although バッグ is not an example of sequential voicing, as it consists of a single morpheme loaned from English, if you apply a similar set of rules to loanwords in Japanese, you discover something interesting. If we applied Lyman's law, we'd find that the /k/ in ばっく could not voice to /g/ because of the voiced obstruent /b/ in the beginning of the word. And when Lyman's law would prevent voicing in native compounding, it manifests as optional devoicing in loanwords. That is, the /g/ in バッグ optionally devoices to /k/, giving us バック. Other examples include:
ビッグ → ビック
ベッド → ベット
This is a native Japanese phonetic rule appearing as a tendency in contexts where it should not strictly apply. This was first pointed out by Kohei Nishimura's thesis, Lyman's Law in Loanwords (2003) and has been the subject of some recent research. In particular, Shigeto Kawahara illustrates in Aspects of Japanese loanword devoicing (2011) that this devoicing is more likely to happen in words that are used more frequently.
(Of course, I think in published material it would still be considered an error, but I think there's at least a possibility that it's a more interesting kind of error than a simple misprint.)