I'm after a scientific position
Do you count math as science?
You have fundamentally misunderstood how PageRank is calculated.
But I am concerned in doing so, some of the pages are going to lose rankings.
When a page is assigned Page-Rank, that is the rank of the page, link out from that page will not have any negative impact on that page's Page-Rank. Say you have a page and it's page rank is a perfect 1. (measuring page rank in a range of 0 to 1). Adding one link to the page would then funnel that perfect score to the next page, but add two links and now the score is divided by 2. But the source page's Page Rank is not affected.
What confuses me in your question is that you are asking about adding links to these high ranking pages, from the home page and other areas of the site. These are in bound links, thus any page rank assigned to those other pages is currently not flowing to the pages in question, so adding links may in fact boost the rank of the pages.
If these are truly orphaned pages and there also no outbound links, then Page-Rank ends in a dead-end, which should in theory break the algo. To overcome this, Google adds in a "teleport" factor that assumes a probability that the user will go to another site entirely without the use of a link. Given that there are no other links, essentially the Page-Rank is wasted, so it would be preferable for you to ensure that you link "outbound" from the high ranking pages to a other relevant and important area of your site, in order to funnel the Page-Rank, instead of waste it.
Essentially the only way to decrease the page rank of a page would be block or break the inbound links to that page, any other action would have no impact the page's own Page Rank.
Now the above is all fine and dandy, but it is really unclear how importance Page-Rank has an ranking, it certainly still counts for a lot, but other factors can have influence as well. So basically there is really no way to know for sure. The best approach would be to take subset of these pages, make the changes an see what happens, if the outcome is positive proceed with the next batch.
Just remember correlation != causation, and Google is always changing things, one can never be certain that the action taken is the true cause of the measured change.