I do not know why the bands are the wrong size but they are faint and there is a lot of primer dimer ( p-d is a small molecule and cannot trap much ethidium bromide so when you can see it there is plenty ).Usually pcrs that work well have little p-d and pcrs that do not work have lots as the enzyme looks for something to do. I think that I would try using a hot start enzyme and run the pcr in presence of 1M betaine in case the low amplification is caused by incomplete melting of the template and if the results still look the same with good amplification change to a new aliquot of primers in case a degraded primer is causing non specific amplification

Before you are going to do the multiplex PCR, always should check the number of primer that are used and concentration of DNA , primer concentration, melting temperature of all primers and extension time .You have to optimize the all primers at same annealing temperature. For that, you have to do gradient PCR for each primer and take the minimum average temperature for annealing. I think number of primer that is used also play a role in getting appropriate bands.

Try to validate your primer sequence specificity and good annealing temperature Tm.