It's not hard to be confused. I have found a number of errors in all those compiled sources -- Rankin, FAG, and the BPC catalog. My advice would be to base your conclusions on the original source documents.

I believe a number of Rankin's errors came from interviewing family members, and also from not realizing some things, like some families living near BPC were Quaker and therefore would not have attended BPC. He had the interest, but wasn't very skillful in genealogy research. And he took the information others gave him at face value, and didn't confirm it with research. He apparently was a trusting soul.

The man named George Donnell buried at Buffalo is listed in the Donnell-authored catalog of BPC as having a stone that says "consort of Ann Donnell who departed this life October 12, 1842 aged 73 yrs." (page 21) That grave is listed in all 3 surveys cited by the catalog of that cemetery. The other listing of interest, on page 74, is for Lydia Ann (Forbis) Donnell, 1771-1844. This is in a section of the catalog that is for no existing stones at the time of the catalog's preparation, and instead is based on records of Rev. Rankin and others. So there may be an error in those records, and the information is on shaky ground consequently.

The Guilford marriage bond compilation done by GCGS has annotations added by members, and sometimes people don't understand that those annotations are not based on the original records, but on someone else's conclusions. That "Lydia Forbis" note is just that kind of thing. The bond itself does not have that information. It lists George Donnel as planning to marry Ann Forbush (another spelling variation of the name that is found in Guilford records). That bond is dated 19 Jan 1796, and the bondsman was Latham Donnell.

There are other marriage bonds for other George Donnell's in the family:
George Donnell & Isabella Kerr, 6 Aug 1784, David Karr bondsman
George Donnell & Anne McClain, 25 May 1815, Joseph McLain bondsman

And not every marriage required a marriage bond. If the couple chose to have the banns called, there might be no surviving record short of indirect evidence such as a daughter's married name in her father's will.

There are two early wills in Guilford for men named Arthur Forbis, one being the son of the other. One is Will Book A: 0116, dated 2 Jul 1780, probated Aug 1781. This is the father, the one who was wounded at the Battle of Guilford Courthouse and who died several days later of his wounds. He names wife Elizabeth, son John, son Arthur, daughters Ann, Elizabeth, Mary, & Marthy (Martha). They were minors at the time.

The will you cite (A:0119) was dated 10 Apr 1789 and probated 1794. It names the beneficiaries you name, and the executors were step-sons John & Robert Rankin. Since a few years passed before he died, any existing estate records may throw a little more light on the family.

The will of David Kerr (made 19 Nov 1802, probated May 1804) lists a son-in-law George Donnell.

George was a name used often in the Donnell family.

The Forbis family, which is one of my own, as well as the Donnell family, can be a challenge. I descend from the Forbis family and also from the marriage of Thomas Donnell & Jane Latham.

I have not worked on the Arthur line related to my own, but supposedly the younger Arthur's wife was named Lydia.