So the first call got you a Mart object that points to the archive site for GRCh37. You can now use that to do things, and you didn't need to do anything more.

What the second call does is figure out which archive you want, based on the version argument. You could use that, but it's not necessary because the first one worked. But for pedantic reasons, please note that 95 is always something in R (because it's a number), but GRCh37 isn't, because unless you put that in quotes, R thinks you want an object in the global workspace called GRCh37 that doesn't exist - it's not a thing.

> 95
[1] 95
> GRCh37
Error: object 'GRCh37' not found

So if you say 'version = 95', that will work because 95 is a number and by definition exists, and R will happily use it. Basically, under the hood what happens is listEnsemblArchives is called, and then whatever you used for the version argument is matched to the 'version' column of the data.frame that is output by listEnsemblArchives. This works because 95 exists, and will be coerced to character when used for matching. But since GRCh37 isn't an existing object, you get the error you see. In other words:

> d.f <- data.frame(A = letters, B = c("GRCh37", 1:25))
## this works because 19 will be coerced to "19" and matched
> d.f[match(19, d.f$B),]
   A  B
20 t 19
#note that everything in the second column is a character
> d.f$B
 [1] "GRCh37" "1"      "2"      "3"      "4"      "5"      "6"      "7"     
 [9] "8"      "9"      "10"     "11"     "12"     "13"     "14"     "15"    
[17] "16"     "17"     "18"     "19"     "20"     "21"     "22"     "23"    
[25] "24"     "25"

## here's the error you get    
> d.f[match(GRCh37, d.f$B),]
Error in match(GRCh37, d.f$B) : object 'GRCh37' not found
## and this is how you would make it work, by using version = "GRCh37"
> d.f[match("GRCh37", d.f$B),]
  A      B
1 a GRCh37

Make sense?