have a look at the "sp_modifystats" system procedure to help remove skew from the "symbol" column stats. I suspect that there is some kind of "default" value for "symbol" that makes up most of the distribution of values in that column which drives up the "total density" for that column. The optdiag utility will show you what the column level stats are for column "symbol". If you post that output here we can see if indeed that is the problem.
This is common. A default hardcoded value is used for most of the rows like "unknown" or blanks (" "). When you search on the column, you typically search on a very specific value, but if you use a variable for that search, the optimizer uses the total density (rather than the histogram of known values when you use the constant literal search string) to determine how many rows will qualify.
Also, given that your index contains many other columns (contributing to longer key length), it could also be the case that a separate index only on "symbol" might be more help. That would require some testing on your part and since you didn't post sp_help output, i can't tell if that might be a viable option - just something to consider.