See Wiki article: Numeric precision in Microsoft Excel

Basically, Excel is limited to an arbitrary number of significant digits. When you multiply two decimal numbers the result has as many more significant digits than the first two. Both formulas start with Logs, which are very likely to have many significant digits.

The first formula merely adds the square of the logs, so probably only gains 1 or 2 significant digits and will probably be accurate to at least 10 digits.

The second formula performs chain-multiplication, X = (((((L1*L2)*L3)*L4)*L5),,,). The result of each of those operations is rounded at an arbitrary digit location. After a hundred or so operations the cumulative error starts creeping up towards a more significant location in the result.

Contrary to the Wiki article, I did not see an increase in presision when using Variant type variables instead of Long types. YMMV