Bandwidth measurements are not accurate

The bandwidth gap between server and client is around 10% .

# ./ethr -s -4

Ethr: Comprehensive Network Performance Measurement Tool (Version: v1.0.0)
Maintainer: Pankaj Garg (ipankajg @ LinkedIn | GitHub | Gmail | Twitter)

-----------------------------------------------------------
Accepting IP version: ipv4
Listening on port 8888 for TCP & UDP
-----------------------------------------------------------
[RemoteAddress]  Proto   Bits/s   Conn/s    Pkt/s   Latency
[    10.3.9.21]    TCP  265.09M        1     --        --
[    10.3.9.21]    TCP    9.60G        0     --        --
[    10.3.9.21]    TCP    9.07G        0     --        --
[    10.3.9.21]    TCP    9.09G        0     --        --
[    10.3.9.21]    TCP    9.66G        0     --        --
[    10.3.9.21]    TCP    9.48G        0     --        --
[    10.3.9.21]    TCP    9.47G        0     --        --
[    10.3.9.21]    TCP    9.61G        0     --        --
[    10.3.9.21]    TCP    9.19G        0     --        --
[    10.3.9.21]    TCP    9.16G        0     --        --
[    10.3.9.21]    TCP    9.55G        0     --        --

./ethr -c 10.3.9.14

Ethr: Comprehensive Network Performance Measurement Tool (Version: v1.0.0)
Maintainer: Pankaj Garg (ipankajg @ LinkedIn | GitHub | Gmail | Twitter)

Using destination: 10.3.9.14, ip: 10.3.9.14, port: 8888
[  5] local 10.3.9.21 port 35617 connected to 10.3.9.14 port 8888
- - - - - - - - - - - - - - - - - - - - - - -
[  ID ]   Protocol    Interval      Bits/s
[    5]     TCP      000-001 sec     8.73G
[    5]     TCP      001-002 sec     8.34G
[    5]     TCP      002-003 sec     8.33G
[    5]     TCP      003-004 sec     8.64G
[    5]     TCP      004-005 sec     8.55G
[    5]     TCP      005-006 sec     8.56G
[    5]     TCP      006-007 sec     8.61G
[    5]     TCP      007-008 sec     8.43G
[    5]     TCP      008-009 sec     8.37G
[    5]     TCP      009-010 sec     8.57G
Ethr done, duration: 10s.
Hint: Use -d parameter to change duration of the test.

Reproduce the report using the shown ./ethr -s -4 server and ./ethr -c 10.3.9.14 client commands, then compare the server and client Bits/s readings. The issue names no source files or tests, so first trace the TCP bandwidth measurement path and determine which reading is expected before defining a fix.