I thought this was a geek heavy site and that networking would be somewhat understood.
In "lay man"'s terms, cellular standards before LTE (I am talking mostly of 3GPP standards but I assume this applies to WiMAX as well) are really really messy.
A normal UMTS radio access network which has evolved from GSM to UMTS might be amix of TDM, ATM och IP based networks with legacy protocols such as SS7. Through that together with general telecom wierdness where much intelligence is deep inside the network.
These networks are optimizing for running voice traffic, so when you are running VoIP you are pushing what the networks are capably of. Also, since there are usually 3-4 nodes between your phone and the gateway to the packet core you will have latency issues since the network is not putting your VoIP-session in a high QoS class.
I am not really sure if you are interested in knowing the gory details, but that was a quick overview. There should be RF- and core-network engineers here on /. who can explain these things more in detail, I only have experience from R&D of these systems in a lab environment.