I played with that idea a little and it does work. See the attached drawing.
I created the KEYNOTE multileader style and set it current. I then created the RED keynote reference blocks (named Keynote Block) which have two attributes; NUMBER and TEXT. I then used the ML command to place a multileader of the KEYNOTE style and when placing it, for the attribute value I right-clicked and used Insert Field to place an Object field that referenced the appropriate Keynote Reference block. After placing a few I created a date extraction that extracted data from the Keynote Blocks. It worked like a charm. If you want to edit the reference number or text, change the Keynote Block attribute. If the number is changed, the multileaders linked to the edited keynote block will all update with a regen. The numbers and text will all update by right-clicking the table and updating the data extraction. Note that Autodesk would not let me attach the data extraction file type (.dxe) so you will have to create your own in order for the updating to work in the table. I left the table in place as an example, though.
I would add a few thoughts. Inserting the object field is a pain. I would probably look into using a LISP for doing so. Short of that, create the field and then copy the keynote for any other iterations. I used two digits for the number for a reason. The data extraction sorts by the number ascending in the way one would sort alphabetically. That means that if you don't have the leading zero for numbers 1-9 then it will sort 10-19 before 2, etc... Adding the leading zero fixes it. Data extraction could also be used on the multi-leaders to see if you have any of a particular value. This could come in useful if you decide to eliminate a keynote from the legend and want to track down any of the multi-leaders for deletion.
