Why doesn't Cisco account for the FCS?

Some context here: https://www.cisco.com/c/en/us/td/docs/ios-xml/ios/qos_plcshp/configuration/xe-17/qos-plcshp-xe-17-book/qos-plcshp-ether-ohead-actg.html

I have a 50mbit circuit with an ISP (L2 MPLS) but the bearer is 100mbit. We pay for the 'L2 rate' apparently.

I get that there's ethernet overhead that needs added but I would class it in three ways:

• Layer 1 stuff: 8 bytes of preamble to sync coms, 1 byte of Start of frame delimited and 12 'bytes' of idle state between transmissions:

• Layer 2 stuff: Source/Dest Mac, Ethertype and FCS

• Layer 3 stuff: Frame payload e.g. the L3 packet.

So when they say they shape at the L2 rate I expect that my QoS policy can ignore 20 bytes per packet (L1 stuff above) and therefore can be built accounting for the L2 overheads only but when i look at what cisco accounts for by default i get this:

• Source/Dest mac = 12 bytes
• Ethertype = 2 bytes
• FCS = 4 bytes

So 18 bytes of overhead exist for each packet. But the Cisco documentation says that they only account for Source/dest mac and Ethertype - So 14-bytes in my scenario.

So why did Cisco ignore FCS in their calculations?