ftDNA and GEDMatch giving different results

[msfarrar]

having uploaded my ancestrydna raw data to ftDNA, my second highest match was to  a gentleman named Bill with    shared centimorgans=68,Longest Block 19 Predicted 2nd-4th Cousin

I made contact and it turns out he is the first cousin of Dick who is the husband of the lady I made contact with. Several of Bill and Dicks relatives have taken DNA tests which are all uploaded to GEDMatch. Comparing them to me on GEDmatch

Bill who is Dicks 1st Cousin:  9.8 on chromosome 9
Dick                   3.6 on chromosome 3
Dicks Sister Judy      3.7 on chromosome 3    6.6 on chromosome 6
Dicks first son ,      0
Dicks second son.      3.6 on chromosome 3
Dicks daughter.        3.6 on chromosome 3

So two questions
1) ftDNA and GEDMatch have significantly different analysis between  Bill and me ?  Which one to believe??   
2)  Does this cluster of low cm matches to several members of the same family add any creedence to me being related to them? 

[hurworth]

Are you saying that FamilyTreeDNA calculates you and Bill share 68cM with a longest shared segment of 19cM, but at Gedmatch all you share is 9cM on Chromosome 9?   That doesn't sound right.   I've seen small variations between FtDNA and Gedmatch but nothing of this magnitude.

I wonder whether she mixed up some kits when they were uploaded to Gedmatch, and Bill's kit isn't there at all?

My feeling is that you may not be related to Dick's side of Bill's family, but you can't rule it out entirely.

[msfarrar]

Yes that is what I am saying. I will double. Check the kit no is correct.
So matches with several family members at >3<4 cm is not a sign I am related ?

[Eric Hatfield]

Experts I have seen tend to say that matches below about 10 cM are not very meaningful - that is why FTDNA's cutoff is about 7 cM. Amounts like 3 or 4 cM may come from so far back that they will be in common with many people who come from the same region.

But I agree with hurworth, it seems very strange that one place says 68 and the other says 9. After all, Gedmatch hasn't done a separate test, but is just using the data uploaded to it, presumably the same data that FTDNA has?

Like hurworth says, I'd check everything, and if necessary upload everything to Gedmatch again, very careful to ensure each upload is correctly identified.