Given your comment that it's that it's stranded then i'd suggest going with two plugs and a coupler. If you want to run high speeds then getting a coupler that is actually cat5e rated rather than a generic RJ45 coupler with no internal impedance control is probablly a good idea.
Punch down connections are only suitable for solid cable. Soldering and heatshrinking is possible but will create an impedance bump which may cause issues at higher speeds. The jelly connectors that telephone companies use will likely create an even bigger impedance bump than carefully applied solder/heatshrink and are also unisuitable for stranded cable.
There is a lot more to making a working UTP cable than just having eight wires connected straight through.
You must meet the required measurements, such as frequency, insertion loss, NEXT, PSNEXT, FEXT, ELFEXT, PSELFEXT, return loss, propagation delay, delay skew, balance, longitudinal conversion transfer loss, etc. Splitting pairs (you seem to do this correctly), untwisting too much at the terminator, exceeding the bend radius, etc. can all cause poor cable performance.
If this is horizontal cable, it needs to be <= 90 meters of solid-core cable with matching terminators. If it is a patch cable, it needs to be <= 5 meters, and it should be stranded cable, with matching terminators. The terminators must match the cable type (stranded or solid core) and be the same category as the cable.
The symptoms you describe usually happen with 1000BASE-T and poor cable that can't pass the test suite. If your company doesn't own the proper tester, it can rent one by the hour.
Best Answer
First a quick note on terminology. The Ethernet standards don't use the terms "hub" and "switch". What the market calls a hub is a multi-port repeater according to the standards and what the market calls a switch is a multi-port bridge according to the standards.
10BASE5 and 10BASE2 coaxial Ethernet used CSMA/CD to manage collisions on the coaxial cable segment. Repeaters could be used to join multiple segments. Repeaters were dumb low level devices, they had no packet buffers and were designed to ensure that if a collision happened on any connected segment it would be seen on every connected segment. This meant detecting collisions and when a collision was detected (either with the repeater or otherwise) on one port a "JAM signal" would be output to the other ports.
Bridges on the other hand operate at a higher level. They work with complete frames and have no need to repeat collisions between ports.
10BASE-T and 100BASE-TX have separate pairs for transmit and receive. With 1000BASE-T things get more complicated but fundamentally there is still a separate data channel in each direction.
These standards can operate in two modes known as "half duplex" and "full duplex". In half duplex mode the twisted pair cable essentially acts as a virtual coaxial cable. If transmission and reception happen at the same time then it is treated a collision even though no signals collided on the wire. The Ethernet controllers use CSMA/CD just as they would on a coaxial cable. Repeaters (hubs) forward collisions to other segments just as they did with coaxial Ethernet.
In full duplex mode transmission is allowed to happen in both directions at the same time and CSMA/CD is not used. Repeaters (hubs) cannot be used on links running in full duplex mode because they would have no way of handling conflicting transmissions.
It is important that both ends of a link are in the same duplex mode. If one end is set to full duplex while the other end is set to half duplex there will be massive packet loss and essentially unusable network performance. Nowadays this is normally handled through auto-negotiation but the system isn't perfect. In particular there are likely to be problems if one end is set to forced full duplex and the other end is set to autonegotiate.
On a modern network the vast majority of links should be running in full duplex mode. CSMA/CD should normally only be used on links to legacy equipment.