You also need to look at the data in the error events memtest86+ has returned, I forget the exact terminology they use but you will have a column of data stored and a column of results returned, if there is an error then these should be different, but how are they different ? is it a specific bit or group of bits which are always high or low.
If multiple sticks are causing errors in any socket and you have eliminated what Adam said (always worth a bios reset to factory defaults as well) then depending on your specific CPU architecture you may be looking at a Northbridge or CPU fault, an overheating northbridge or maybe just a CPU which needs reseating or checking for bent pins.
Another thing to remember is that you can get confusion in where the error is in terms of offset compared to the hardware if bank interleave is switched on.
Finally the Memtest86+ Author has been very helpful in the past when I have had results back which were confusing. But I think false positives with memtest86+ are very rare.
Check the offset between the expected and error and the start and finish addresses of where you get error data to work out if it is just one or a few address lines. if it is you most likely have a hardware fault.