netspeed reports larger speeds than normal. For instance, while using
Firefox to download a file, or apt to upgrade the system; the application
(Firefox or apt) would report for example 350 KB/sec, while netspeed would
report 575 KB/sec.

I realise that there's a discrepancy between the data amount in application
terms and the data in raw interface terms. However, I also checked with
iptraf and while it also reports greater speeds than the applications, it is
only a minor increase: 370 KB/sec when the application reports 350.

The problem appears at any level of speed, not just large speeds. Examples:
application: 5, iptraf: 6, netspeed: 8.5
application: 10, iptraf: 11, netspeed: 20
application: 20, iptraf: 23, netspeed: 35
application: 50, iptraf: 55, netspeed: 85
application: 100, iptraf: 113, netspeed: 170

These are just examples, because the speed reported by netspeed also seems
to fluctuate a lot, but generally it stays around 70% more than what the
application reports, while iptraf only adds about 10%.

The same applies to upload speeds.

Changing the update interval does not fix the problem or improve it.

        Hi,

 I believe netspeed relies on the interface byte counters which in the
 case of Ethernet will count Ethernet frames, even broadcast ones.  If
 you're on a busy LAN with Windows host, you'll typically experience a
 non-null amount of traffic.

 Could you try measuring traffic with wireshark and see whether
 netspeed's traffic matches the wireshark Ethernet counters?

   Thanks,

I'm afraid I'm not used to WireShark and I haven't been able to figure
out how to do that. Can you point me to some help?

 Install Wireshark, capture all data on your interface, stop capture,
 and see Statistics > Summary.