Barring additional unstated details, this pattern should work:
([a-z])([^\.])\n([a-z])
\1\2 \3
Your current pattern is capturing the second-to-last character of the line with ([a-z]), and then the last character in the line with [^\.] -- but since you didn't put that into a capture group, it will be discarded.
You might consider using
([a-zA-Z0-9\s])([^\.])\n([a-zA-Z0-9\s])
or even just
([^\.])\n
replaced with
\1
(space at the end of that replace pattern).
I plumped up the search string to include capital letters and digits, you may or may not want to do that. And I'm not sure if/why you really need the first capture group at all -- is it important that the second-to-last character and the next-line first character be a letter (or number or space)?
But at the very least, putting a capture group around that character class that excludes the period should give you what you need.
--Kerri