Using a USB/ethernet adapter is of course adding "another thing" with it's own complexities. USB 1 & 2 stand no change of running at gigibit. You'd need USB 3 (in both the port and the adapter) and of course the host bus and OS would need to be capable of driving it that hard.
A subtlety that is often overlooked when people get into the data networking "numbers game" is that the equipment cites performance in a metric called "Link Rate" (or DataLink Rate or Layer 2 Rate) whereas when people "talk" about "speed" and try to measure it using Speedtest and the like, they are talking about something else called "throughput." The two are both expressed in "bits per second" but they are not at all the same thing, in the same way that instantaneous "velocity" and "average speed" of a car are both expressed in km/h but they also are not the same thing.
Data networking Link Rates are not some theoretical measure of "capacity" (even WIkipedia gets this wrong,) they are a statement of "time" expressed as a ratio. A link rate expresses how long it takes to transmit 1 bit, not how many bits can be transmitted in an arbitrarily chosen unit of time.
Thusly, a link rate of 10 bps mean "it takes one tenth of a second to transmit 1 bit" NOT "you can send 10 bits every second." It just so happens that we choose to express it as a ratio based on a prticular unit of time (whole seconds.)
Virtual all data network technologies do not transmit continuously, they are "bursty." Some bits get sent, then silence, then some more bits, then silence and so on. If you were to run a "speed test" you see an average value, which factors in the the silent periods, Link Rate does the opposite and only expresses the time taken to transmit 1 bit when those bits happen to be in flight.
The time to transmit 1 bit is a precise measure - which is why it's always been used "in the business" - averages (such as computed by speed test) are subject to the vagaries and variations causes by different length bursts, different length silent periods and so on which make them much more variable. Not to mention the various "protocol overheads" which differ between technologies and traffic competition which all make "speed test" "indicative" at best.
So amongst data networking professionals data networking Link Rates are something we use more or less as a "label" rather and then anything we sweat over in order to ensure link X is "just so." We use speed tests (in the knowledge of all the protocol overheads and other factors) as a sort of "wet finger" test to ensure a link is in the right ballpark, but again we don't stress over the numbers.
So, for example, being orders of magnitude out would get our attention, but, let's say, speedtesting a 100mbps ethernet link at 90mbps would not be getting us excited as it's "about right."
To test your ISP link, you need to remove as many other complicating factors as possible, hence the guys are advising you to use something we know is very reliable and efficient (ethernet) and avoid using WI-FI, powerlines, USB/ethernet adapters and so on. Such are complicating the investigation as they introduce their own variances. And the more of them we have in the test path, the more "things" there are that could be complicating factors.
It's also worth mentioning that not all routers are equal. IP routing is a relatively simple but not trivial computing task and any router has to be "sized" to be able to cope with the amount of traffic it expects to handle. Routers effectively have RAM and a CPU inside them, like a PC, and they need to be "fast" enough to cope with the anticipated load. So a router "sized" for little 8mbps ADSL link probably would not have a horsepower to cope with a 200mbps FTTC link and so on. (The metric is called "WAN-To-LAN Routing Capacity.") This is why ISP's often send you a new router when you upgrade your package - your old one may not have the grunt for the upgraded line.
A US web site called SmallNetBuilder often tests router WAN-to-LAN routing capacity when he tests kit. You might take a look there and see if he's ever tested your router and what results he got.