That's not quite what I said yesterday. GEDmatch won't necessarily provide a more accurate picture, but examining the matches on GEDmatch might help to explain why there is an apparent discrepancy.Like Peter has suggests viewing the results on GEDmatch is really the only way to see the true picture.
Other research into matches shared by parents and their children has suggested that around one-third of DNA matches are false matches. But it doesn't matter, provided you follow the strategies in my Masterclass, and focus on the matches whose online trees suggest a connection. (The chance of getting a false match with someone who really is a cousin is pretty small.)
When you have tens of thousands of matches there is little point puzzling over matches where there is no obvious connection (unless the amount of DNA shared is so high that it can't be ignored, or the match is shared with a cousin). I wasted years playing with Chromosome Browsers before I saw the light and tested with Ancestry. I wrote the Masterclass because I didn't want LostCousins members to make the mistake that I did!