Although @Leo Laohoo and @Elliot Dierksen touch on an issue, often overlooked, i.e. many Cisco routers are not capable of running all their interfaces (or sometimes even one interface) at full wire-rate/line-speed (sustained). The OP has an ASR 1000 Series label and the description of the issue appears that more than 100 Mbps cannot be obtained by any one flow, including noting five concurrent flows were able to obtain 500 Mbps. From that data, one might assume a single flow should be able to at least also obtain 500 Mbps.
A quick check, of the router's CPU consumption, such as hitting 100% during such a test, would be good indicator that the capacity limit of the router has been reached. Conversely, if the CPU's usage is, say, below about 80%, would normally indicate a different issue.
Reading the other replies, @balaji.bandi raises an interesting possibility, i.e. your version of iPerf might have a bug.
@MHM Cisco World suggests decreasing MTU. Something you might try, although, again, I would first try an UDP test as it avoids TCP issues. However, as a side note, since you didn't describe the kind of WAN being used, as with an MTU issue, something might be happening along the path.
Lastly, @Flavio Miranda asks if you're shaping, which is a good question, although "normal" shapers are used to set an overall bandwidth limit, not per flow. I recall a Cisco Catalyst 6K supports per flow policers, so again, might it be possible there's something in the path?
BTW, my mention of BDP and LFN considerations are also per flow. No guarantee either is your root cause, but they can cause behavior as your OP describes.
