gfdeconv not the same as LFSR?

I have been trying to implement the following formula in matlab:

b84x84+...+b1x+b0 = Rf(x)g(x)[ bn-1xn-1+...+b85x85 ] + g(x)

Where the result is an 85 bit (84 downto 0) vector of "check bits", the part in brackets on the right [ bn-1xn-1+...+b85x85 ] is the data we are sending.

I believe: Rf(x)g(x)[ bn-1xn-1+...+b85x85 ] is forming a CRC where f(x)g(x) is the CRC polynomial.

What I think the formula is saying is to find the remainder of the data divided by f(x)g(x) and then add g(x). This is all using finite field GF(2).

To do this I was performing the following:

FxGx = gfconv(fs_x,gs_x);
[q, R] = gfdeconv(data,FxGx); % so R is our CRC I think
check_bits = gfadd(R,gs_x);

This would then give me my 85 check bits to append to the data prior to transmission. However, the result wasn't correct.

I replaced the [q, R] = gfdeconv(data,FxGx) line with my own LFSR using FxGx as the feedback terms for the LFSR and now I get the expected result.

I was under the impression that gfdeconv was effectively doing the LFSR for me, but I guess it's doing something slightly different. So my question is, is gfdeconv not the same as LFSR?