1) I don't see a BGP configuration inside the VRF on the PEs. they have to have an eBGP session set up between the VRF on the PE and the CE so they can peer.
2) you'll want your IBGP session to be setup with local-address of the loopbacks instead of the physical interfaces. it might not affect much in this simple lab setup with only 2 PEs but I believe would be good to maintain best practice even in a lab environment so you don't get bit later in an actual deployment.
I actually have a similar lab setup using some SRX's to teach myself some mpls and mpls-application stuff (when I can find free time from customers!). i think logical tunnel interfaces (lt-0/0/0) work on the J-series like they do on the SRX (at least on 10.1), so you could actualy do some interesting stuff with a full mesh square with your 4 J-series acting as a service provider cloud and then have a virtual-router that eBGP peers over a logical tunnel interface to a service provider VRF on the same physical box. it complicates the config a little, but it lets you do some more stuff with less lab gear, especially with L3 VPNs where you can add multiple lo0 subinterfaces to VRs or VRFs to simulate routes going across the mpls coud to other CEs.
I can post up a config example if you like.
good luck! i'm right there with you trying to figure this stuff out lol 🙂
I concur with winclemd. Based on the config output in your pastebin, your L3VPN is working correctly. You just need to setup the PE-CE BGP session under the [ routing-instances VPN protocols bgp ] stanza.
And yes, iBGP should always be loopback-to-loopback, unless you have a really solid reason to not do it.