Hi Guys,
I feel like I'm missing something very obvious.
I'm doing the last bullet in 2.2 which tells me to "ensure R5 prefers to reach prefixes learned from BB2 via RIP and not BGP".
R5 learns 3 prefixes from BB2 both by RIP and BGP. Also it has eBGP peerings to R1 and R2. If I modify the AD on R5 as per SG, the prefixes learned by eBGP are still sent to R1 and R2 despite being marked as "RIB-failure" in BGP table. So R1 and R2 will prefer to reach these prefixes via BGP.
Since redistribution is done from routing table these prefixes will not be even redistributed to EIGRP domain and there is no risk of any loop.
Where did I make a mistake? :-)