I do not know the solution to your problem... And, what exactly the problem is? Is it that the highlight is word-wise? And you would prefer have it one continuous interrupted highlight?
You mentioned problem with missing spaces in annotations exported to text file. It is an old problem which Onyx tries to handle (they say it is caused by a bug in Adobe pdf library...), but no success so far. If I am not mistaken it has to do something with text coding used. If I remember well most/all pdf generated from LaTex have this problem. It seems to me that UNICODE text coding is the source of problem.
If you want to experiment, try to recode pdf from UNICODE to older coding.
Also the highlight behaviour might be caused by this.
It seems to me, that Onyx programmers could solve at least the "highlight text fragmentation" behaviour. They could look for fragmented highlight text and join it to single one. Try to ask Booxstor to add it to TODO list.
It is hard to say how much programming work it is. It may be one to few hours, it may be much, much more, depending on how the things are implemented in software.