Verdict on Mann

I know that Stuart Mann's "An Indo-European Comparative Dictionary" is
regarded as disreputable, unreliable, and often erroneous. I want to
know, to what degree -- extremely? slightly? moderately? I especially
would like to know regarding his distribution of PIE *k^ vs. *k -
although I have heard it said many times that in reconstructions of
PIE, *k^ greatly outnumbers *k, in Mann's dictionary *k^ appears
initially 414 times (275 times before vowels) whereas *k appears
initially 1028 times (687 times before vowels) -- not to mention *g^
initially 167 times (134 times before vowels) vs. *g initially 396
times (187 times before vowels) and *g^h initially 131 times (103
times before vowels) vs. *gh initially 331 times (204 times before
vowels). The point is, in his reconstructions the velars are
considerably more frequent than the velars, as one might expect in a
natural distribution in a language. But is Mann's data to be rejected
wholesale? Or can I use it at all reliably? (If anyone can tell me
the occurrences of velars vs. palatals in LIV for comparison, I would
greatly appreciate it.)

Andrew