Coding and non-coding transcripts overlap in databases, am I doing something wrong?

[thejustpark]

While working with the UCSC and RefSeq noncoding data extracted as instructed, I found a strange problem.

First of all, I collected non-coding transcripts from RefSeq and UCSC annotations using the filters "NR_" and "cdsStart=cdsEnd" in UCSC Genome browser and coding transcripts using the filters "NM_" and "cdsStart!=cdsEnd".
I would expect that a transcript cannot be both protein-coding and non-coding, but there are many transcripts appearing both in coding and non-coding annotations.
For example, I found that NR_02784 and NM_013796 refer to the same transcript.

This trend is also found between UCSC and RefSeq annotations.
For example, UCSC lncRNA 'uc009ura.2' is exactly the same as RefSeq protein-coding transcript NM_027327.
How can this happen?
Does this mean I somehow collected the data in the wrong way or is it their annotation problem?
In the case of latter, what can I do about these transcripts? do I have to throw out all such transcripts of my considerations?

Thanks guys for your time and helps.

HJ.

[kmcarr]

For example, I found that NR_02784[sic] and NM_013796 refer to the same transcript.

I am assuming you mean NR_027804 and here is the explanation from the corresponding GeneID entry at NCBI:

NR_027804.1 RNA Sequence
Status: VALIDATED
Description
Transcript Variant: This variant (2) includes an alternate internal segment, compared to variant 1. This variant is represented as non-coding because the use of the 5'-most supported translational start codon, as used in variant 1, renders the transcript a candidate for nonsense-mediated mRNA decay (NMD).

[thejustpark]

Thank you kmcarr,

Yes I meant NR_027804. Sorry for the typo.
Thanks for the check on NR_027804.
It seems some transcripts can work as either protein-coding or non-coding.
I should have checked the webpage myself.

Again, thanks a lot.