you can do it by reading the last 6 caractes as noted but it is a bad idea since you will instead lose the ability to distinguish between tapes differing only in the first two positions such as 123456L5 and 223456L5.
If you ever get a robot that only report the first six digits (e.g. Oracle) you have a problem. It's not consider "best practice" to have duplicate ID's in a barcode range - not even using the density type as part of the barcode.
Using the Density portion of the barcode only really would affect you if you had more than 9,999 tapes in any one density. Other than that as long as your library is able to differentiate between the first and last six digits used to find the tape it is up to you. Its really a matter of preference as long as those two conditions are meet. However if you are close to 10,000 media in a density I would recomend a different numbering system as in the long term it may cause problems.