I assumed the readers would follow the link to see the original results. GMP is undoubtedly the state-of-the-art of arbitrary-precision arithmetic. The point of this table was to compare bignum implementations provided by different languages, so programs that use GMP bindings are excluded. Probably it would make sense to include the best result that uses GMP, but I think GHC is quite representative of GMP's performance.
Amazing that Wilfred Hughes and yourself took such a straightforward approach instead-of complaining that a lot of the programs used GMP. (Take that as praise.)
If we don't want the programs that used the GMP library, then we can simply not show those results!
Except "Some languages rely on GMP to implement built-in bignums. They are marked with an asterisk (*)." — why are they shown?
(And why isn't there a C program + GMP measurement?)