A few questions that might help narrow down the problem you were seeing:
How exactly did you test the fail over?
How much time did you wait for things to stabilize before deciding the fail-over did not work and turning the second connection back on?
How is your outbound routing setup? Default routes or full tables? If defaults, it would be helpful to see any static routes that might be present.
Assuming that 19094 is still announcing the aggregate, the problem of filtering should be a non-issue (assuming they don't filter the 701 path from their upstreams). In any case, things seem to look ok from an outside perspective to most everyone who has commented.