Hello Evgeny
I discussed this over with @ReneMolenaar and he said that the lesson may be somewhat confusing. Specifically, R3, as you correctly mentioned, plays the role of a multicast router as well as a multicast client. In the same way, R1 is a multicast source, and a multicast router. He said he would revise the content by adding a multicast source connected to R1 and a multicast client connected to R3 so these roles will be separate.
Now having said that, looking at your comments, we can say the following:
Yes, this is correct. If R3 was a host, the permit rule would not be needed. The permit rule is used so that R2 can find the DR which is R3.
This too is correct, since R2 is itself the DR it does not need the entry in the ACL to communicate with the DR.
The ip pim passive command will cause an interface to not send out or accept any PIM messages from other routers. The router will instead consider that it is the only PIM router on the network, and thus act as the DR. This command will actually cause PIM neighborships to not form, so the whole multicast topology will malfunction. @ReneMolenaar labbed it up and the topology did indeed fail. This command should only be used when there is only a single multicast router on the network. If there are more than one multicast routers, and you want to filter out PIM messages, use the ip pim neighbor-filter command instead. You can find out more about the passive functionality of PIM at the following Cisco command reference.
I hope this has been helpful!
Laz