Um.
The first allows you complete control over the power consumption of your system. An MFSU has a maximum output EU/t, so the sum of all your consumption cannot exceed this.
The second one is probably a bad attempt at creating a backup storage system, where:
(a) You have an excess production of EU compared to your consumption
(b) and/or your EU production is continuous whereas your consumption is not and vice versa
What you would need to do is to throw a transformer between the line connecting the producer and the consumer, right below the MFSU. If you want to increase the throughput of that line, you need add more transformers in parallel. The second system is a lot more useful for when you have multiple systems producing and consuming EU. This will allow you to have a variable EU throughput that regulates itself.
Though it is technically possible with the first set up, the second one is more redundant. My system uses a combination of the first and the second to regulate the flow of EU in my system. Keeping in mind that it regulates EU in two ways:
(1) If your production/consumption are not equal or continuous, fluctuation in these rates will be handled by the "backup" MFSU storage (i.e. A buffer)
(2) The error in the second set up shown in the image is called a feedback loop. Though it is different than how it works with signals, the principle is the same. The MFSU is charging itself. Though, I do not know if the programming would cause the MFSU to send an EU packet to itself. i.e. When EU is requested from the MFSU, does the MFSU acknowledge the state of the MFSU as non-full and would send a packet through the loop back to itself and the consumer, simultaneously (since it is a valid path in the wiring). To avoid all problems with this, directing EU flow with transformers or additional MFSUs is appropriate.