This is most likely an unfortunate side-effect to Cisco's non-standard inline power scheme.
This is a case where the port timing is indeterminate. Some Cisco 'Inline Power' systems send illegal FLPs (they advertise that they are something other than an ethernet, or 802.3 device). When CableIQ sees these, it cannot determine anything about the switch (or even that it is a switch). Therefore, we call it an 'illegal port'. We would get the same response if connected to an 802.5 token ring port (on UTP).
The reason that you don't see this often is that Cisco switches will 'play fair' if they see a legal advertisement from another device. The Cable IQ sends legal advertisements to get the Cisco switches to send legal ones back. Unfortunately, we cannot wait forever to see if the switch is going to send legal FLPs back; we wait a long timeout (several seconds) then give up and call the port what it is - illegal. Cisco does not send any particular illegal value - it sends different values on each port, and changes them at irregular intervals.
Most managed switches have internal priority settings - it's more important to route packets than it is to handle an FLP. Therefore, if the switch is busy with other tasks (management, high traffic levels, etc) the time it takes to handling an FLP may become very long.