...but it DOES work!
If you were to determine the letters as each occupying a centimetre, then, as I have also indicated, the Golden mean for a line 22 centimetres long would fall at approximately 13.6, hence in the fourteenth position.
This is not, however, how the general sense for the Golden Proportion (outside of geometrical constructions) were carried out. Rather, precisely the importance of the Fibonacci series in whole numbers (and whether by that name or not), is of greater reflective significance. Hence also why Filipas points out that using the Fool in the measurement makes one 'lose' the connection.
Even if constructing relatively large structures one may gain an appreciation for the proximity to the Golden Mean and usefulness of the numbers one, thirteen and twenty-one (and hence, to their respective first, thirteenth, and twenty-first letters) by constructing squares (or even a golden spiral) within a Golden Rectangle 17cm (ie, 34/2 - or even more importantly, in terms of our considerations, (21+13)/2cm) wide and its accurately drawn height, and another within a rectangle 17cm wide by 27.5 cm (55/2) high (those measurements come out quite easily if one uses 5mm graphing paper). It will be very difficult indeed to visually perceive the difference between the geometrically determined Golden Mean of the line of 21/2, and the one which was constructed with the defined measures using the graph paper/ruler (at the thirteenth position).
The connection, however, remains, as First, as Thirteenth (not fourteenth), and as Twenty-first.
This is not diminishing the role of the Fool: rather, it requires that it be placed last... and UNNUMBERED. The card cannot be placed in first position, for it would then occupy the space between (on the ruler) 0 and 1, ie, the first numbered space, which is the province of the first numbered card - the Bateleur, or, if using the Alef-Beit, of Alef.
Part of the possible difficulties in these is precisely how one is going to determine the Golden Mean of a series.
If one uses the method applied by kwaw - a method which is also correct from a strict linear geometrical construct, then the Golden Mean of the twenty-two centimetres will be at approximately 13.6, and therefore fall in the fourteenth partition (hence the assumed Nun). This, however, contradicts the sense in which numbers and considerations of Golden Proportions are normally met. In kwaw's considerations, a card numbered zero would have to assume the space occuring before the space between 0 & 1 - ie, a negative number (between 0 & -1), and outside of the range of letters. Indeed the overall considerations lend support to the un-numbering of the Fool as final card.
_____
In terms of the Tetraktys, this is normally considered to be in the shape of an equilateral triangle.
If designed according to a Golden Triangle, then the full circle is implied by two pentagrammes - precisely for the reasons mentioned by kwaw, in which the apex of each triangular point is of 36 degrees, hence ten points are needed to complete a circle.
If one wants to place these Golden Triangles in tetraktys form, however, then another six (downward pointing ones) are created...