Q2) How do I treat transactions where the base item, discount coupon usage and installation are counted as separate Transaction IDs?

A2) Ideally, for discounts, it makes sense to do some data cleansing beforehand to roll up the info so both base item and discount count as a single transaction so it doesnt skew the analysis. For installation though it might be worth keeping separate if not all customers choose to also pay for the installation. This might justify treating the installation as a separate product/service that should contribute on its own to the Frequency count.

Ultimately though the degree of precision and accuracy goes back to what the RFM is being used for I.e. Is it for Customer Segmentation, to estimate Customer Lifetime Value or to predict future transaction patterns - and even for that are the predictions on an aggregate basis or meant to be specific to individual customers for marketing interventions?