The reason nearly everyone on here will chine in on the use of NOLOCK (locking hints in particular) is primary because of the byproduct of using it altogether (dirty reads/potentially bad). The most common misconception with this is using it makes things run faster (and this isn't the case). It's typically best to let SQL manage it's own locking/escalations.

In looking at your attached execution plan, it looks like you used the Tuning Advisor already (IMHO, take what that thing provides you with a grain of salt). I agree with Sean, the differences in your estimate/actual may be due to out of date statistics. Consider, updating them before running your query or completely rebuild the following indexes:

- MT_20121207_Transactions_forselectforamazon_1

- _dta_index_transaction_shipping_9_464772763__K9

- _dta_index_transaction_shipping_notes_7_660197402__K1_2