While running quality control steps for an RNA-Seq experiment I can see a decrease in the read quality score at the first few nucleotides (see below).
I read the fastQC manual and it said on the Per Base Sequence Quality that poor quality at the start of the run could be caused by bubbles in the flowcell. But with that said the per-tile quality plot looks fine.
Another possibility is that a warn / error is triggered because of a short loss of quality earlier in the run, which then recovers to produce later good quality sequence. This can happen if there is a transient problem with the run (bubbles passing through a flowcell for example). You can normally see this type of error by looking at the per-tile quality plot (if available for your platform). In these cases, trimming is not advisable as it will remove later good sequence, but you might want to consider masking bases during subsequent mapping or assembly.